Methods of Sociological Inquiry
Methods of Sociological Inquiry SOC 357
Popular in Course
Popular in Sociology
This 184 page Class Notes was uploaded by Deron Effertz on Thursday September 17, 2015. The Class Notes belongs to SOC 357 at University of Wisconsin - Madison taught by Staff in Fall. Since its upload, it has received 41 views. For similar materials see /class/205178/soc-357-university-of-wisconsin-madison in Sociology at University of Wisconsin - Madison.
Reviews for Methods of Sociological Inquiry
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/17/15
Sociology 357 Prof Pamela Oliver I Variables amp Units of Analysis A Variables describe units of analysis B A variable is a dimension along which units of analysis differ or vary A variable is a set of exhaustive and mutually exclusive attributes into which all of the units of analysis in question may be classi ed C EWes UNIT OF ANALYSIS VARIABLE CATEGORIES OF VARIABLE individual income exact income to nearest dollar or categories eg lt 10000 10000 24999 25000 34999 etc individual eye color blue brown green hazel etc household size number ofpeople l 2 3 4 5 6 residing together Organization sex composition female percentage to nearest whole percent or categorize lt20 female 2050 female 5180 female gt80 female Census tract average income mean to nearest dollar D Level of measurement 1 nominal Exhaustive amp mutually exclusive categories E g Eye color major sex gender 2 ordinal Nominal ranks Course grade A AB B etc course level E I A 3 Interval Ordinal plus meaningful metric so distance between 1 amp 2 distance between 23 amp 24 Few examples temperature scales 4 Ratio Interval true zero Height in inches income in dollars number people enrolled in a class Notes Nominal ordinal qualitative Can do frequencies percentages proportions mode Interval ratio quantitative Can do qualitative means standard deviations correlations all other statistics Very few statistics especially for ordinal Ordinal variables with 5 categories can usually be assigned numbers and treated as interval II Propositions l A proposition is a statement about variables A Univariate proposition is a statement about one variable at a time quotMost UW students drink beer at least once a weekquot Variable frequency of beer drinking UOA individual UW students Statement quotmostquot drink once a week or more B Bivariate proposition is a statement about the relation between two variables quotMales drink beer more often than femalesquot Variables 1 sex 2 frequency of beer drinking Statement gives relation between them C Multivariate proposition states a complex relation among three or more variables quotAmong nondepressed students males drink beer more often than females but among clinically depressed students males and females drink beer equally oftenquot Variables 1 sex 2 frequency of beer drinking 3 whether clinically depressed or not 2 A Hypothesis is a type of proposition Some use as synonyms I used it to mean the proposition being tested in a particular research project Stern uses it for the finding of the project even if it is an afterthefact result and limits it to bivariate 3 An assumption is a proposition that is taken or assumed to be true within the context of a particular research project You always need to make some assumptions to study something else Important types of assumptions A Measurement assumption The assumption that a particular operationalized variable is an adequate representation of a particular concept All research has measurement assumptions B Theoretical assumptions Assumptions about how things generally work apart from the specific thing being studied III Relationships 1 Qualitative eg A Blacks are more likely to be Democrats than whites are Variables race party choice UOA people Percentage tables 2 Quantitative Positive relation Height is positively correlated with weight On the average taller people weigh more Negative relation The speed with which a task is performed is negatively related to the accuracy with which the task is performed the faster you work the more mistakes you make Curvilinear relation The level of stress is curvilinearly related to age with the highest levels of stress occurring in middle age and lower levels of stress occurring among those younger and older IV General form of a proposition Conceptual For population P in condition C independent variable X causes dependent variable Y Operational For sample p in condition c independent variable x has a statistical association with dependent variable y V Cr quot quot quot defining the r J you will use for deciding what category of a variable each unit of analysis belongs in Sampling Why Sample Why not study everyone Debate about Census vs sampling Problems in Sampling What problems do you know about What issues are you aware of What questions do you have Key Sampling Concepts Who do you want to The Theoretical generalize to Population What population can The Study you get access to Population How can you get The Sampling access to them Frame Who is in your study H The Sample Copyright 2002 William MK Trochim All Rights Reserved Sampling Process Units of Analysis people Target Population Population of Interest Actual Population to Which Generalizations Are Made De nedListed by Sampling Frame List or Procedure Sampling Frame List or Rule De ning the Population A Generalization Target Sample Response Rate Sample The people actually studied Method of selection V I List of Target Sample Key Ideas Distinction between the population of interest and the actual population de ned by the sampling frame Generalizations can be made only to the actual population Understand crucial role of the sampling frame Sampling Frame The list or procedure de ning the POPULATION From which the sample will be drawn Distinguish sampling frame from sample Examples Telephone book Voter list Random digit dialing Essential for probability sampling but can be de ned for nonprobability sampling Types of Samples V Probability Simple Random quot Systematic Random Strati ed Random Random Cluster Strati ed Cluster V Complex Multistage Random various kinds NonProbability Quota V Convenience PUFPOSIVG Probability Samples A probability sample is one in which each element of the population has a known nonzero probability of selection Not a probability sample of some elements of population cannot be selected have zero probability Not a probability sample if probabilities of selection are not known Probability Sampling Cannot guarantee representativeness on all traits of interest A sampling plan with known statistical properties Permits statements like The probability is 99 that the true population correlation falls between 46 and 56 Sampling Frame is Crucial in Probability Sampling If the sampling frame is a poor fit to the population of interest random sampling from that frame cannot fix the problem The sampling frame is nonrandomly chosen Elements not in the sampling frame have zero probability of selection Generalizations can be made ONLY to the actual population de ned by the sampling frame Types of Probability Samples Simple Random Sampling Each element in the population has an equal probability of selection AND each combination of elements has an equal probability of selection Names drawn out of a hat Random numbers to select elements from an ordered list Strati ed Random Sampling1 Divide population into groups that differ in important ways Basis for grouping must be known before sampling Select random sample from Within each group Strati ed Random Sampling2 For a given sample size reduces error compared to simple random sampling IF the groups are different from each other Tradeoff between the cost of doing the stratification and smaller sample size needed 0 for same error Probabilities of selection may be different for different groups as long as they are known Oversampling small groups improves inter group comparisons Systematic Random Sampling1 Each element has an equal probability of selection but combinations of elements have different probabilities Population size N desired sample size n sampling interval kNn Randomly select a number j between 1 and k sample element j and then every kth element thereafter jk j2k etc Example N64 n8 k6488 Random j3 Systematic Random Sampling2 9 Has same error rate as simple random sample if the list is in random or haphazard order Provides the bene ts of implicit strati cation if the list is grouped Systematic Random Sampling3 Runs the risk of error if periodicity in the list matches the sampling interval This is rare In this example every 4 11 element is red and red never gets sampled If j had been 4 or 8 ONLY reds would be sampled Random Cluster Sampling 1 Done correctly this is a form of random sampling Population is divided into groups usually geographic or organizational Some of the groups are randomly chosen In pure cluster sampling Whole cluster is sampled In simple multistage cluster there is random sampling within each randomly chosen cluster Random Cluster Sampling 2 5 a Population is divided into groups Some of the groups are randomly selected For given sample size a cluster sample has more error than a simple random sample Cost savings of clustering may permit larger sample Error is smaller if the clusters are similar to each other Random Cluster Samplng 3 Cluster sampling has very high error if the clusters are different from each other Cluster sampling is NOT desirable if the clusters are different It IS random sampling you randomly choose the clusters But you will tend to omit some kinds of subjects Strati ed Cluster Sampling Reduce the error in cluster sampling by creating strata ta 3 of clusters Sample one cluster from each stratum The costsavings of clustering with the error reduction of strati cation Strati cation vs Clustering Strati cation Clustering Divide population into Divide population into groups different from each comparable groups other sexes races ages schools cities Sample randomly from Randomly sample some of each group the groups Less error compared to More error compared to simple random simple random More expensive to obtain Reduces costs to sample strati cation information only some areas or before sampling organizations Strati ed Cluster Sampling Combines elements of strati cation and clustering First you de ne the clusters Then you group the clusters into strata of clusters putting similar clusters together in a stratum Then you randomly pick one or more cluster from each of the strata of clusters Then you sample the subjects Within the sampled clusters either all the subjects or a simple random sample of them Multistage Probability Samples 1 Large national probability samples involve several stages of strati ed cluster sampling The Whole country is divided into geographic clusters metropolitan and rural Some large metropolitan areas are selected with certainty certainty is a nonzero probability Other areas are formed into strata of areas e g middlesized cities rural counties clusters are selected randomly from these strata Multistage Probability Samples 2 Within each sampled area the clusters are de ned and the process is repeated perhaps several times until blocks or telephone exchanges are selected At the last step households and individuals within household are randomly selected Random samples make multiple callbacks to people not at home The Problem of NonResponse 1 You can randomly pick elements from sampling frame and use them to randomly select people But you cannot make people respond Nonresponse destroys the generalizeability of the sample You are generalizing to people who are Willing to respond to surveys If response is 90 or so not so bad But if it is 50 this is a serious problem The Problem of NonResponse 2 Multiple callbacks are essential for trying to reduce nonresponse bias Samples Without callbacks have high bias cannot really be considered random samples Response rates have been falling It is very difficult to get above a 60 response rate You do the best you can and try to estimate the effect of the error by getting as much information as possible about the predictors of nonresponse Nonprobability Samples Convenience Purposive Quota Convenience Sample Subjects selected because it is easy to access them No reason tied to purposes of research Students in your class people on State Street friends Purposive Samples Subjects selected for a good reason tied to purposes of research Small samples lt 30 not large enough for power of probability sampling Nature of research requires small sample Choose subjects with appropriate variability in What you are studying Hardtoget populations that cannot be found through screening general population Quota Sampling Preplan number of subjects in speci ed categories eg 100 men 100 women In uncontrolled quota sampling the subjects chosen for those categories are a convenience sample selected any way the interviewer chooses In controlled quota sampling restrictions are imposed to limit interviewer s choice No callbacks or other features to eliminate convenience factors in sample selection Quota Vs Strati ed Sampling In Strati ed Sampling 0 In Quota Sampling selection of subject is interviewer selects random Callbacks rst available subject are used to get that who meets criteria is particular subject a convenience sample Stratified sampling Highly controlled quota without callbacks may sampling uses probability not in practice be much sampling down to the last different from quota block or telephone sampling exchange But you should know the difference for the test Sample Size Heterogeneity need larger sample to study more diverse population Desired precision need larger sample to get smaller error Sampling design smaller if strati ed larger if cluster Nature of analysis complex multivariate statistics need larger samples Accuracy of sample depends upon sample size not ratio of sample to population Sampling in Practice Often a nonrandom selection of basic sampling frame city organization etc Fit between sampling frame and research goals must be evaluated Sampling frame as a concept is relevant to all kinds of research including nonprobability Nonprobability sampling means you cannot generalize beyond the sample Probability sampling means you can generalize to the population defined by the sampling frame Sociology 357 Methods of Sociological Inquiry Lectures Notes 2 Basic Concepm Units of Analysis 0 The objecm we study 0 People 0 Families 0 Cities 0 Newspaper articles 0 Classes in school Variables 0 Dimensions or aspecm of units of analysis which vary Variables MUST vary 0 Formal definition of a variable is a set of exhaustive and mutually exclusive categories 7 Every unit of analysis must fall into exactly one category ofavaria e 0 Variables are defined by researchers Examples of Variables or categorize Level of Measurement I Nominal Exhaustive amp mutually exclusive categories Eg Eye color major sexgen er Ordinal Nominal ranks Course grade A AB B etc39 course level E I A I Interval Ordinal plus meaningful metric so distance between 1 amp 2 distance between 23 amp 24 Few examples temperature scales Ratio Interval true zero Height in inches income in dollars number people enrolled in a class Qualitative Quantitative I Qualitative Nominal ordinal qualitative Can do frequencies percentages proportions mode Quantitative Interval ratio Can do qualitative m ans st dard deviations correlations other statistics I Very few statistics especially for ordinal Ordinal variables with 5 categories can usu y e assigned numbers and treated as interval Propositions 0 A Erogosition is a statement about variables 0 A univariate proposition is a statement about one variable at a time quotMost UW students drink beer at least once a week Variable frequency of beer drinking UOA individual UW students Statement quotmostquot drink once a week or more Bivariate Proposition 0 A Erogosition is a statement about variables 0 A bivariate proposition is a statement about the relation between two variables quotMales drink beer more often than femalesquot Variables 1 sex 2 frequency of beer drinking Statement gives relation between them Multivariate Proposition 0 A Erogosition is a statement about variables 0 A multivariate proposition states a complex relation among ee or more variables quotAmong nondepressed studenm males drink beer more often than females but among clinically depressed studenm males and females drink beer equally often Variables 1 sex 2 frequency of beer drinking 3 whether clinically depressed or not Hypothesis I A hypothesis is a type ofproposition I Some use proposition and hypothesis as synonyms I I use hypothesis to mean the proposition being tested in a particular research project This is the most common usage I Some use hypothesis to mean a proposition Whose this unce 39n I Stern uses hypothesis for the bivariate nding of a project even if it is an a erthefact result General Form of a Hypothesis I Conceptual For population P in condition C independent variable X causes dependent variable Y I Operational For sample p in condition 0 independent variable X has a statistical association with dependent variable y Qualitative Relations Used with qualitative variables Need to be stated in words listing which categories of one variable have more or fewer unis of analysis in each category of the other variable EX Blacks are more likely to be Democrats than whites are Variables race party choice P ercentag es Quantitative Relationships 1 I etween quantitative variables I Positive When variable one is greater the other tends to be too Height is positively correlated With Weight On the average taller people Weigh more Waght Hagat Quantitative Relationships 2 I Negative When one variable is greater the other tends to be smaller The speed of performing a task is negatively related to accuracy On the average the faster you Work the more mistakes you Number of errors Time m complete task Quantitative Relations 3 I A curvilinear relation can be any nonlinear relation but is especially a relationship that is rst positive then negative or vice vers a Educaunn Age There is a curvilinear relation between age and education in the US education rises With age but because of historical increases in education rates older adults have less education than younger adults OperationalizationMeasurement I To operationalize a variable is to say how you will measure it I To measure a variable is to use specific observational or operational procedures I The operationalization of a concept is the same thing as its measure or measurement I This has two parts Operationalization Part I I First the procedures for collecting data eg observe ask questions I Question How o en have you smoked marijuana vs Have you every smoked marijuana I Observation of dgew Take motion picture count frames in which position changes vs observe face to face count number of times hands touch head Operationalization Part II I Exact distinctions among categories of variables within a procedure I If counting how you tell the beginning and end of countable things I If distinguishing among types of actions or characteristics must develop rules for an exhaustive and mutually exclusive set of different types Operational Variables are Created by Researchers I No measured variable is natural All have to be created by the decisions of the researcher But some are easier to operationalize than others I Researcher makes sure the categories are exhaustive and mutually exclusive I Researcher decides how precise to be Precision vs Accuracy I A more precise variable makes ner distinctions 7 Height in inches instead offeet e Shading of eye color greyblue sky blue deep blue violetblue bluegreen pure green yellowgreen light brown dark brown etc instead ofbroad groups blue brown hazel I Accuracy is correct classi cation into the category I Tradeoff between precision and accuracy Harder to be accurate With ner distinctions Range I Categories must be exhaustive so must encompass the full range the subjects exhibit I Other over 100000 and not applicable are ways to make a variable exhaustive I Categories should be defined so only a few subjects end up in the residual category Indicators Indicators are indirect measures of Variables Not really operationalization of the thing itself but something known to be highly correlated with it eg Where there s smoke there s re Smoke is an indicator of e Church attendance is an indicator of religiosity You might attend regularly for crass nonreligious reasons or not attend due to practical constraints but there is a high correlation between frequency of churc attendance and subjective religiosity Indicators are important for concepw Which are dif cult to measure directly BEGINNING ISSUES for Methods I Why bother to do research Why bother to be scienti c What is scienti c anyway How do values t in A Wh do we do research Learn how to do it We want to answer questions Singleton Ch 1 gives examples of a variety of research projects all about helping B List topics that interest us Problem is to make them research able by way of nding the empirical questions within the topics The factual issues as distinguished from values judgments policies Scienti c questions can be answered by observation Stern quotquestion of factquot quotempirical questionquot A statement you can try to con rm or discon rm by looking at the evidence ofthe senses or sensing technology Rules 1 it must be possible for it to be true or false NOT de nitions eg bachelor is unmarried man NOT assumptions treated as tautologies e g quothuman behavior is self interestedquot if quotinterest is de ned to include anything you want including wanting to be self sacri cial Assumption of selfinterest in economics and much of sociology has this tautological character you assume selfinterest and infer interests from behavior But you could make selfinterest empirical statement something like this ask people to say what goals are important to them and what actions they believe will help them obtain these goals then give them choices among actions they have the capacity to perform and see if they take the actions that would be predicted from their statements C Examples 1 The grading system is bad value statement How turn into empirical Possibilities Grading process lowers students39 self esteem Grading process hinders leaming For each of these would further re ne to exactly what you mean So that you could nd out whether the statement is true or false 2 quotRacism is a problem at UWquot What does this mean What are empirical statements relevant to it Statement about subjective feelings of whites and blacks Statement about incidence rates Marwell argues that incidents have always been there that there are changes in whether people just suffer them silently or get together and protest Statement about objective levels of discrimination Statement about general ignorance of history economics of race relations 3 quotSexism is a problem at UWquot Again issue must be turned into empirical statements Rates of aggressive sexuality Frequency of professors comments students39 subjective feelings frequency of tensions and arguments among women and men D Role of values Values what ought to be quotWhite students should be less prejudiced toward minoritiesquot quotLecture sections in methods should be limited to 25 studentsquot vs quotMost white students have prejudiced attitudes about blacksquot or quotInstructors teaching methods would prefer their sections to be limited to 25quot E Evidence Stern Relates to Singleton knowledge as precise description concepts need agreedupon meanings l unsupported assertion 2 appeal to authority Note sometimes this is acceptable e g if professor is giving lecture or if the source is somebody you have reason to trust But you are doing no thinking or checking for yourself If it really matters you would ask the authority what evidence shehe has either verbally or by looking up the article 3 casual observation unconcretized abstractions not just list examples but state rules we can apply to new cases give an operational de nition quotMen are usually more aggressive in class than womenquot What did you actually see Men punching other men on the nose Men talking more than women Voice intonations of men and women being different Women using more selfdeprecating or apologetic speech Were there a few men who were much more quotaggressivequot than any woman or were all the men more aggressive than all the women In fact you have reason to worry that the person telling you this hasn t thought the matter through and might be lumping together a wide variety of unsystematic and even selective observations F Singleton s characteristics of science a empiricism look at senses not just re ected b objectivity people can agree on what they say truth does not depend on who is doing the observing When 6 my daughter believed in ghosts and said that only those with ghost detecting ability could see ghosts Her statement put ghosts outside the realm of science c control procedures to eliminate major sources of bias and error e g eliminate selective observation overgeneralizing inaccurate observation Inaccurate observation Most of us are not good observers unless we are careful Jane Piliavin reports using a social psych text that in fact alternates pronouns so that exactly half of all references are to quothequot and half are to quotshequot But one student complained that the book was biased and always talked about she Low status people e g women blacks report having the things they say in business meetings attributed to others People often report observing things that just are not true The scientific remedy is careful conscious observation according to welldefined rules 2 Overgeneralization You are correct about what you saw but assume incorrectly that what you saw in a few applies to many Newspaper reporters talk to six people at a demonstration and characterized everybody that way You talk to your two teenage kids about quotyouth todayquot and assume you know all there is to know For that matter you assume that what you know to be true of your own friends is true of everybody Similarly there is a tendency to overgeneralize from the negative behavior of one person to a whole group this is what we mean by stereotyping and it always makes us angry when we hear it done about our own group 3 Selective Observation You notice the things that prove your belief and ignore the others If you re a woman who objects to what you think of as aggressive men you might get angry at all men failing to notice that most of the noise is actually being made by very few men and that most men don t act that way You confine your research about teenage sex to those who come up to a booth labeled quottell me your experiences with teenage sexquot Statements like quotYou39re not like other X39squot or quotexception that proves the rulequot are signs of selective observation G Reliability different observers use abstraction in same way agree about what they see Validity the operational measure is quotreallyquot what you think it is link between concept and measure We will come back to these H Theory Gives definitions of concepts assumptions describing circumstances undre which they apply set of interconnected abstract principles and propositions How and why the empirical generalizations are true General principles that explain many empirical generalizations Provide either understanding of causal processes or subjective understanding of people s motives 1 Science involves BOTH research and theory in constant interaction with each other RM Old idea discontent causes SM eg rises and falls in movements due to shifts in discontent But data shows measures of discontent don t correlate with action unless you use the action to measure the discontent which is a tautology and not scientific Next RM says external resources are the factor specifically claim that civil rights movement was started by the resources of white liberals Data McAdam time line year by year on protests and white foundation money show money followed the protests clearly refuted the hypothesis new understanding is internal resource political openings theory of cycles about how protest draws in external money leads to professionalization II Scienti c Attitude I see this as scholarly attitude not a matter of qualitative vs quantitative or humanities vs hard sciences A Knowing is better than not knowing knowledge is worth seeking for its own sake Not introspection Youth looks out into a mriror and can see only him or herself Maturity looks through a window and sees the world and other people as they are In this sense science is mature It is not about knowing yourslef better and not about deciding what you think others ought to do it is about knowing more about something outside yourself about understanding what other people think ought to be C A scientist respects the facts respects the evidence May sometimes be very resistant 39 39 act I 39 quotJ if it is quot J 39 39 quoty possible to discoutn it but is generally interested in knowing what all the facts are and what the broad range of evidence is like Would rather know an unpleasant truth than not D A scientist tells the truth about his or her own research insofar as he or she is able Doesn t hide the facts doesn39t make things up Humans are humans and nearly everyone tries to put the best possible construction on his or her work and everyone knows this so you learn to read between the lines In particular if something is not clear you can probably assume the worst rather than the best But a scientist never lies about the resarch no matter what You don t lie about the sample size you don t lie about the measures you don t lie abouat the statistical results That is one of the few things that can get you fired from being a tenured professor Discuss cheating brie y in this light E A scientist knows what she knows and knows what she does not know The scientist knows where the evidence is and where there is no evidence The scientist knows what his or her range of expertise is and doesn39t try to cast an aura of authority around everything he or she says Of course the scientist may know that he or she has studied something well enough to know about it even when it is outside his or her official expertise NOT SUMMER 1 Go over Stern Ch 2 pp 5660 s l 6 7 variables relation results IV DV Asking Questions A few pointers MultipleChoice Likert Format Strongly Strongly Disagree Agree Neutral Disagree Agree 1 2 3 4 5 Never Rarely Sometimes Often Always 1 2 3 4 5 Match response to item 0 Frequency NeverAll the time 0 Likert Scaling DisagreeAgree 0 Quality PoorExcellent 0 Service Not WellExtremely Well 1 Use Simple Sentences 0 No double negatives 0 Eliminate vagueness poorly de ned terms 0 Avoid objectionableIrrelevant questions Avoid Double Negatives It is not the case that I have never cheated on my tax returns Never should one not help others The likelihood of depression recurring after the discontinuation of psychotropic drug treatment is greater than if drug treatment is never used as part of therapy Avoid Vague or Ambiguous Teims How many times in the past year have you talked with a doctor about your health Is health care easily accessible for your family Studying accounts for a majority of the activities I do at college Tests are stressful I relax by using drugs yr 4 Avoid Objectionable or Irrelevant Questions NV How old are you Have you ever tested positive for HIV virus Have you answered each question truthfully How many years of education were you able to complete I believe crack is one of the four food groups Avoid Doublebarrel Questions Is your doctor friendly and reasonably priced Were your caregivers courteous and friendly I am often fidgeting and on edge I find that I am more attentive and remember more if I have eaten before a study session Balance questionsresponses How was the service at this hospital Excellent Very Good Great Are you depressed frequently Sad is the best descriptor of me right now My depressed mood keeps me from doing fun things Reverse Score to Reduce Response Bias I am often sad Strongly Agree Undecided Disagree Strongly Agree Disagree 4 3 2 I often feel happy Strongly Agree Undecided Disagree Strongly Agree 39 agree 2 3 4 Exhaustive amp mutually exclusive categories What is your age under 10 1020 2030 3040 4050 How did you last travel to the supermarket car bus foot walking public transportatio ya What is your marital status single married divorced Even vs Odd categories Strongly Agree Disagree Strongly Agree Disagree Strongly Agree Neutral Disagree Strongly gree Disagree Slide 1 Slide 2 Slide 3 Doing the Observation Exercise Tips Methodological Concepts in Ass ignm ens Ob sa39v anon 7 opeiaannaiinng a vanaisie 7 iniensuisieeave reliability Experiment 7 isniaang causal ieiaanns by ennanuing exaanenus vanahles 7 Lu e afiananmizaann Questionnaire 7 opeiaannaiinng a vanahle Wins iniiiapie inneainis 7 csinsaueivaiiaicy relanuns zmung different measures Examples of Things to Study People in public places TV shows Radio shows or ads Magazine ads News articles from newspapers or electronic archives Slide 4 Slide 5 Slide 6 PreliminaIy Unstructured Obs ervation paitners can do this separately does not have to be at exactly the sarne tirne or on the sarne subjects But agree on the type ofobject you are observing or type ofplace you are observing people in You may observe together ifyou wish but write your observations and ideas for research separately before discussing thern Sampling Choices u u i units ofanalysis You must end up with at least 30 cases units of analysis Two options 7 1 You can sarnpie by a rule in which the rule is operationaiiZing the sarnpie toward making the pmiect nonrobvious case de ning and counts trade for observation or agree whom to observe in a given time period Option 1 Sampling by Rule Don t cheat create arule and then follow it independent1y Examples observe everyone who comes through the door Watch all commercials on channel 7 between 7 and 8 pm Use keyword Search to nd all articles about murder in May of2002 Slide 7 Slide 8 Slide 9 Option 2 Pick Subjects Directly To save time when you cannot observe together you may each choose a sample of 15 documents 6 g magazme ads observe and code the variable f n trade samples In field observation you may agree whlch subiect The observatlons ofthe variables mustbe 100 end r l indep entryou mustnotpeek atthe othe s coding or discuss it in any way Testing Reliability You must observe the same subjects independently without checking each other s wor Ifyou sampled by rule calculate sampling error and locate those subjects you both observ ed Cal ulate the percentage error disagreement in coding the dependent variable Calculating Reliability 1 N total number of distinct people seen by both or elther Sl sample difference 1 number oftimes partner 1 observed a subiect partner 2 did not see sz sample difference 2 number oftimes partner 2 observed a subiect partner 1 did not see A number you agree on same subject same variable co e c coding difference number oftimes you observed the same subiectbut coded the variable differently Slide 10 Slide 11 Slide 12 Calculating Reliability 2 Coding Error CE CAC proportion you both saw that you disagree about in the variable Sample selec on error SE s1 s2N proportion oftotal cas es that one person saw but not the other Evaluating Reliability Perfect reliability is the goal zero errors s assignment need to do a variable that is complex enough thatthis is not easy Even 10 error is falrly high for reliability Try to understand the source of all errors and how they could be avoided Ife or pr rr is low discuss whatyou did well in the ocedures to produce low error Conditional Percentages 1 Dependent variable is qualitative 2 Crosstabulate the data 3 Calculate percentages for the dependent 39 39n each category ofthe E a o 1 o 0 a 3 lt E T 4 Compare the percentages across categories ofthe independent variable Slide 13 Slide 14 Slide 15 Crosstabulate the Data Dependmt Calculate Cell Row Column Totals Calculate Conditional Percentages lndep endent Slide 16 Slide 17 Slide 18 Final Table Interpretation Males bit 53 othe time 24 othe licked 59 othe time compared to 33 formales a percentage difference of 26 Other was onl w omen Conditional Means 1 Dependent variable is quantitative 2 List the values for the dependent variable separately for each category ofthe independent van e 3 Calculate the mean for the dependent variable separately foreach category H 39 A r A variable a t independent variable List Dependent Variable Scores by Independent Variable Males 13 17 De might variable 22 23 25 44 34 m ale traumatic 1U 23 13 55 33 31 33 27 3m 16 22 42 Na zn number ufmales Np 27 numberuffemales megs sumufscures are m3 sum ufscures Slide 19 Slide 20 Slide 21 Chlu ztelvbrs quratelyfcr chGrch 1127 rumba Cfferrda E291093smufm 1ferrd Nn20 rnmber drmles Edens42sinndm hrnales Mean EbgnNn 5422Z7 10 Final Table Men a o n 271 405 N 20 27 Interpretation Women took 13 4 seconds longer than men on average to complete their transactions Hypothesis Testing Conrirnis statistical association in the direction predicted which is fairly strong Discon rms opposite direction from prediction OR zero association when o predicted anonzero statistical association Indeterminate statistical association is in the direction you predicted but is weak Slide 22 Testing Prediction of Zero Association This is harder to falsify Only an exactly zero association con nns A large association discon rms A weak association is indeterminate Soc 357 Fall 2006 Sampling Theory Exercise Due Tuesday October 19 or earlier in class Reading Glicken quotSamplingquot and Lecture notes This exercise is worth 4 of your grade It is a small quotthoughtquot exercise to verify that you understand the basic sampling concepts that are otherwise not tested in this course You will show your understanding of these concepts by giving simple examples of them Do NOT write long involved explanations one or two sentences examples for ead r term are plenty In your examples of specific sample types be sure to indicate whether choices at eadi step are made randomly or nonrandomly SUBMIT TWO 2 COPIES OF THIS EXERCISE I will return the original with comments and save the photocopy This exercise may be redone to raise your grade but you will have to invent new examples I will also confirm that you have not merely copied the examples of someone else who previously received a better grade In developing your examples it is ok to use one possible theoretical population eg Elementary sd rool students in Madison for the whole exercise and structure your answers in terms of different possible ways to sample from that population 1 Give an example that clearly shows the meaning of and relationship among the population of theoretical or substantive interest the sampling frame the sample and the actual population to which generalizations may be made 12 points 2 Give examples that clearly illustrate the difference between haphazard or accidental sampling and random or probability sampling 6 points 3 Give examples of simple random sampling and systematic random sampling from the same population or sampling frame emphasizing their similarities and differences 6 points 4 Give related examples that clearly illustrate the similarities and differences between stratified random sampling and random cluster sampling including a discussion of the circumstances for which ead r is appropriate 6 points 5 Give an example of multistage stratified cluster sampling which clearly illustrates the key features of this approadq and when it is appropriate 6 points 6 Give an example of simple quota sampling that clearly illustrates how it differs from a related example of stratified random sampling 6 points 7 Give an example of a nonrandom purposive sample for a case in whid r a purposive sample would be appropriate and contrast it with an example of a mere convenience sample 6 points Learning Through Observation Soc 357 Summer 2006 A Basic Distinction Structured Observation observation guided by clear rules about who you will observe your sample and what you will observe about them Unstructured Observation observation that does not use systematic control techniques 7 aka ethnography participantobservation eld research What is Field Research Observation of naturally occurring events Goal isto build scienti c theory ie To generate hypotheses AKA participant observation and ethnography Roots in American Sociology The Chicago School 1920s 1930s Areas of study ethnic groups and other small communities deviance amp powerlessness occupations amp professions ordinary life When is it useful Obtaining insider s view of events Maintaining complexity of actions events Studying relationship between person and setting different units of analysis Projects where other research methods are precluded by money ethics other methodological problems When there is little known about the phenomenon Steps Choosing a site Sampling Access amp SelfPresentation Recording Observations Analyzing Data Choosing a site Lo and amp Lo and Start where you are depends on where you are 7 Eg Adler ampAdler Wheeling and Dealingquot Sampling Nonrandom selection of people amp events Goal Maximizing variationdifferences Multiple cases more recent y Use theoretical sampling identify analytical categories you think will have important distinctions amp sample based on those Use snowball sampling gather informants based on recommendation of others Example Eliasoph Avoiding Politics 0 Wanted to know why people don t talk about politics le Investigating a cultural phenomenon Chose to study ve groups two volunteer groups one social club two community activist groups Theoretical sampling setting might make a difference to talk 0 Finding similar discourse across all groups that avoided making publicly spirited claimsquot in public places Getting Access amp SelfPresentation Formal organizations often have gatekeepers Reliance on key informants Researcher has to decide what role she will la Participant 9 ParticipantObserver 9 Observer Goal Preserve natural order of group Covert research 7 Eg Tearoom Trade Recording Observations I Primarily unstructured observation Use eld jottings to record incidents Write up eld notes at the end of each day Focus on concrete behavioral details don t analyze Keep analysis feelings impressions in a separate set of notes I Verifiability Analyzing Data Coding for themes highlighting bits of text assigning meaning Developing ideas Drawing and verifying conclusions Use triangulation multiple sources including respondent review Two ways to approach theorizing Grounded Theory Glaser amp Strauss 1967 theory that emerges from observations not theory that is derived a priori 7 aka Inductive reasoning Extended Case Method Manchester School of social anthropology 1950s 1970s Burawoy 1998 start with a theory nd a case that helps you test amp modify your theory 7 Quasideductive reasoning Examples of Theorizing Creating system for classifying behaviors phenomena 7 Eg HumphreysTearoom Trade Looking at particular instances of more general phenomena 7 Eg Browne The Used Car Game Understanding processes 7 Eg Adlers Shi s and Oscillations in Strengths of Field Research Maintains complexity of social phenomena Helpsto develop hypotheses that can be tested systematicall Helps us understand connections between variables Some argue it s more ethical to take into account meaning of phenomena from subjects point of view Limits of Field Research Inef cient method of collecting data Can t descr be distribution of phenomena in a population Limited generalizeability Can t test hypotheses without reference to other cases Merton s critique Making evidence fit your heory rather than other way around Debates Validity of observations 0 On Stagequot Effects gt Deception becoming an inSideri extended observation cross checking Triangulation s gt Biind measurement Warmeup period Protocois for interaction 0 Researcher Bias gt Reflexivi i p g o Incomplete Observations gt insider informants multiple observers thirdrparty observers Debates Ethics Ethics 7 Of covert research 7 Of using people to thher career goals o Adier amp Adier Reciprocity is essentiai Debates Taking on the Worldview of Others Dorothy Smith standpoint Theory Goal should be to elaborate the standpoint of the people we study Pierre Bourdieu amp others researchers concerns aren t the same as those of the people we study Debates Is This Really Science Recall characteristics of Science 7 Developing testable hypotheses 7 Based on observation and veri cation Theories are meaningless if we don t understand motivation for behavior This can be done rigorously findings are valid Analyzing Qualitative Research Adler amp Adler as exemplar 7 Looking at process and variations within that process 7 Theore ical rationale insights into the problems involved in leaving deviancequot 7 How do they use evidence to support their descriptions Evaluating Adler amp Adler Sampling theoretically meaningful Getting access amp roles played Wnat problems did they encounter Recording observations amp Data analysis was it systematic Theorizing was it posthoc Which ndings are most trustworthy Why Which ndings are least trustworthy Why Examples of Observation Projects Primping at the Pub Ice cream cone eating Purchases timing different kinds Analysis of TV commercials Analysis of advertising photographs Homework for Thursday Read Observation Assignment Instructions Brainstorm what kind of project you would like to do 7 Public observationbehavior 7 PrintTV study 7 Gender race age and Some other more complex variable Preparing Your Report Soc 357 Fall 2006 Writing up your research You will write up your ndings separately But you can discuss your results as a team as much as you want you are encouraged to do so You must write a group process report individually Notes on Format amp Writing Please make sure to include the same subject headings listed in the instructions Be as detailed as possible If you made a mistake in the execution of your project you still get full credit if you identify a What you did incorrectly and what were the consequences b What would have been the correct procedure Testing Reliability You must observe the same subjects independently without checking each others work If you sampled by rule calculate sampling error and locate those subjects you both observed Calculate the coding error the percentage of disagreement in coding the dependent variable Calculating Reliability 1 Compile Data Compare two partners data Mark all cases in which you sawthe same person and gave the same variable code with A Agree To measure Sampling Error Put a by a any case one partner saw that the other missed did not code t all To measure Coding Error Mark the cases in which you sawthe same person but gave them a different variable code with D Disagree g Calculating Reliability 2 Counts To calculate Sample Error you need S1 sample difference 1 of s on 1 s data sheet number of times partner1 saw a subject partner 2 did not see 52 sample difference 2 of s on 2 s data sheet To calculate Coding Error you need A number you agree on The number of A s on ode sheet same subject same variable code 1 and 2 should agree on of A s C Coding difference Number of D s on one code sheet number of times you observed the 5 subject but coded the variable differently 1 and 2 should agree on of D s Calculating Reliability 3 Computations o N total people seen by either partner 0 82 0 Sampling error SE 1 2IN proportion oftotal cases that one person saw but not the other Coding error CE CAC proportion you both sawthat you disagree about in the varia Ie erson teams calculate 23 possible pairs for 4 person teams ca culate 2 pairs those that seem most alike and those that seem most different 0 n 2 92 1 Evaluating Reliability Perfect reliability is the goal zero errors But for this assignment need to do a variable that is complex enough that this is not easy Even 10 error is fairly high for reliability Try to understand the source of all errors and how they could be avoided If error is low discuss what you did well in the procedures to produce low error Preparing data for hypothesis testing You have 4 choices Each partner analyzes the data she collected Use the data from the partner you believe was most rate Create a composite data set using the good data from eac partner Do the analysis for each data set You must explain what you did in the report The Logic of Hypothesis Testing You have a hypothesis about the relationship between two variables in your study You examine the data to see if it confirms or disconfirms your hypothesized relationship or if the data are inconclusive Conditional Percentages Dependentvariable is qualitative Crosstabulate the data Calculate percentages forthe dependent variable separately Within each category ofthe independentvariable Compare the percentages across categories of the independent variable M J Crosstabulate the Data Independent Dependent Calculate Cell Row Column Totals Check the row and column totals against the data before proceeding Calculate Conditional Percentages Independent Dependent Divide each cell total by the total for that category ofthe independent variable Final Table It is OK to use the original proportions or to turn them into percents Rounding error Final Table herquot Interpretation Males bit 53 of the time of 26quot was only slightly different for men an women Rounding error Conditional Means 1 Dependent variable is quantitative 2 List the values for the dep endent variable separately for each category of the independent 3 Calculate the mean for the dependent variable separately for each category of the independent vanab e 4 Compare the means acro independent variable 55 categories of the List Dependent Variable Sc Independent Variable ores by Males Dependent Females g variableisnumber 26 44 ofsecondsittook 36 23 82 to complete sales 34 10 24 74 23 10 29 transaction 31 49 57 19 55 69 29 14 39 31 39 55 33 33 27 33 31 30 16 30 47 22 22 26 42 Nm 20 number ofmales xm 542 sum of scores Nf 27 number offemales 2m 1093 sum ofscores Calculate Means Separately for Each Group Females Calculate Means Separately for Each Group a es Male 3 Final Table Men Women Mean Seconds for Transaction 271 405 N 20 27 Interpretation Women took 134 seco nds longerthan men on average to complete their transactions Hypothesis Testing CONFIRMS statistical association in the direction predicted which is fairly strong or statistically signi cant if you are doing a signi cance test DISCONFIRMS opposite direction from prediction OR zero association when you predicted a nonzero statistical association INDETERMINATE statistical association is in the direction you predicted but is weak or not statistically signi cant Testing Prediction of Zero AssOCIation This is harder to falsify Only an exactly zero association con rms A large association disconfirms A weak association is indeterminate Goals for today 0 Discuss ndings 39om preliminary observations 0 Finalize your plan to hand in at the end of class i How you Wlll sam le 2 What yariables you Wlll observe 3 How you Wlll classiry your obseryations Listthe total set of all possible classlflcatlonSoutcomes for each yariable of interest at Wlll you o it you are unsur a out ow to classliy 4 How Wlll you minimize obseryer effects What Wlll you do it this becomes an issue 5 Do you know how to construct a table to record your 0b5eNathrl5 7 Slide 1 Slide 2 Slide 3 Sociology 357 Methods of Sociological Inquin Hypothesis Testing Methodological Concepts in Ass 1gnm ens Ob sa39v anon 7 opersuunshnng a variable 7 intersuhieeuye reliability Exper ent 7 lsulanng causal relanuns hy cuntxullmg exuaneuus variables Questionnaire 7 opersuunshnng a variable with rnuiupie inuieaturs 7 Cunstxuctvalidity relanuns zmung diffa39ent measures Induction Induction is reasomn from the Specific to the genera17Ernpiriea1 geneiniization There is no logical proofofinduction future cases may be different from those you ye seen However sampling theory which we will do later tells us how we can u e population mean is between 22 and 25 Slide 4 Slide 5 Slide 6 D eduction Deduction is reasoning from the general to the speci c following the rules oflogic 7 All men are mortal Socrates t a man 7 Therefore Socrates is mortal Deduction is important in scienti c research for the logic of falsi cation Illogic of Proof of Theory If theory is correct then X is true is rue Therefore theory is correct INVALID LOGIC A irming the co ent X mightbe tme for another reason Illogic of Rejecting Data Because You Reject the Theory If theory is correct then X is true Theory is fals Therefore X is not true INVALID LOGIC Denying the antecedent X can be tme even ifthe theon iswrong about WHY it is true Slide 7 Slide 8 Slide 9 Logic of Falsification Iftheory is correct thenXis true Therefore theory is not correct VALID LOGIC Falsification We carrrrotprove theorres to be correct We CAN roveLheoriesLobelN Research proceeds on a logic offalsrfrcatrorr We suhrect eurrestu testsw cuuld faisrfythe e lfaLheury zvmds falsi canun we sayitis un rmedquot hutpruveh e lfaLheury repeatedly zvmds faisrdcaaurt webmld Bur wrurrgiater C ausation 39L 1 It is generally dif cult or39 r 39 0 an al to Criteria for 39nferring causation from observables r statrstrcai assocratroh two things Vary together 7 Cause precedes effect In trrrre r Extraneous variablesquot are ehrrrrhated as possrbie expiarratrorrs for the reiatrorrshrp We wru study thrs m depth later 7 an r entry the mechahrsm forthe causeeffect reiatrohshrp We know hoW rtw Slide 10 Slide 11 Slide 12 Statistical Association Fornow we will focus on assessing the statistical relation between two variables For qualitative independent and dependent variables we will compare canditienaipercentages For qualitative independent and quantitative d pendentvariabies we will compare conditional ans For quantitative independent and dependent variables ii 4 im r regressions Sex and Ice Cream Cone Eating statistical Association licked 59 othe time compared to 33 formales a percentage difference of 26 Other was onl women D erence ofCond onalPercentages Sex and Time to Complete Sales Transactions N 20 27 Interpretation Women took 13 4 seconds longert an men on average to complete their transactions Difference of Conditional Means Slide 13 Slide 14 Slide 15 Correlations xam le spent and elapsed Lime ofthe transaction is 43 Correlationsm gebetwe n7 erfectnegatlve correlation to 1 perfectpositwe conelation A 250 correlation means there is no monolonlc linear relationshl The strenth ofa correlation nses with its square 7 If cunelanunls 7 Dr V 7 than 7 lfcunelaaunis 9 meg then 81 quhevanancelsexplamed 7 lfcunelaaunis z unz then l4 quhevanancelsexplamed Full Logic onypothesis Testing Research Syllogism f c theory Aquot measurement assumption And ifY measuresindicates B measurement assum Then X will be statistic puon ally associated with Y prediction Andthmeasuresindi ThenX wtll he Data 1 X is statistically assncizmd with Y prediction is conec Cannot prove that A causes B but cnn rms m suppurts ory at A causes B assumptions Confirmation of Theory Research S 1ch ates A measurement assumpaun Bates B measurement assumpann statlsacally assnnatea Wth Y predichun Slide 16 Disconfirmation of Theory stemming unmet r WWMM Xms msm kw A tamem umgtmuymumsmms Murmansan mmun umuseremiwmr mlmgt Data 2 x isNOT statistically associated with v prediction is wrong Then either A does not cause B or x is not a measure ofA or v is not ameasure ofB By logical necessity at least one assumption is wrong Falsi cation ofthe research syllogism BUT Falsification may be in error due to sampling error or extraneous variables 7 later THIS IS AN EXAMPLE OF A BADLY DESIGNED SURVEY Principal Investigator ID No Larry Neuman Dept Sociology SOCIOLOGY RESEARCH METHODS Code QUESTIONNAIRE Instructions Please answer all questions as truthfully and accurately as possible Check or mark the appropriate boxspace for each question or print an answer in blanks provided All answers are confidential and complete anonymity is assured Your participation is voluntary and will help us greatly Thank you Part I Individual Background and Demographic Information Nb l 5 Lquot O 00gt Which is your sex Male What is your current marital status What is your birth order Female What is your class standing in the University Fresh Soph Jr Sr Special How old were you at your last birthday Years What is your maj ors If unknown at this time write Unknown in space provided What is your current GPA for all college courses 38to 400 28 to 299 18 to 199 36 to 379 26 to 279 16 to 179 34 to 359 24 to 259 14 to 159 32 to 339 22 to 239 12 to 139 30 to 319 20 to 219 Below 120 If you do not have any idea of what your GPA is check here Never Married single DivorcedWidowedSeparated Married Do you have any children adopted or natural or spouses Yes I was an only child No Raised in Institution Youngest of Siblings Oldest of siblings In 1Iiddle some siblings older amp younger 9 What race or ethnic heritage do you consider yourself BlackAfroAm ericanNegroid WhiteAngloCaucasian YellowAsianOriental Native AmericanRed BrownHispanicChicano Other 10 What kind of area did you grownup spend high school years in RuralFarm Area VillageSmall Town under 10000 pop Small City 10000 to 50000 Medium CitySuburbs 50000 to 300000 Large City 300000 to 1 million pop Maj or Metropolitan Area Over 1 million 11 Where us your hometown Madison Wisconsin Elsewhere in Wisconsin Elsewhere in Midwest East CoastAtlantic Seaboard West CoastMountain South Southwest Outside USA 12 What is your religious affiliation Protestant Baptist Catholic Lutheran Jewish Quaker Muslim Atheist Part 11 Family Background Questions 13 What is your father s occupation 14 What is your mother s occupation 1f parent is deceased retired or unemployed please state that If parent is homemaker please write Homemaker in the space 15 Do both parents work full time Yes No 16 What social class would you assign to your family based on income amp life style Upper Class 1Liddle Class Upper Middle Class Lower Middle Class Blue CollarWorking Class PoorLower Class 17 About what would you estimate your total family income to be for 1981 before taxes Over 75000 per year Over 50000 but under 75000 Over 30000 but under 50000 Over 20000 but under 30000 Over 10000 but under 20000 Under 10000 per year Part 111 Questions About University Courses 18 Which school or college of the University are you in 19 Have you ever taken a statistics course prior to this semester NO GO TO 23 skip 20 21 22 YES 20 1F YES What department was it in 21 What was the number of the course 22 Why did you take it Check all that apply Curious interested Required for my degreemajor Friends recommended it AdvisorProf recommended it Wanted to avoid more difficult stat Course 23 Math Anxiety is an emotional fear or great discomfort that some people feel when confronted by complex symbols and number manipulation Do you think that you have math anxiety No SKIP to 25 YES 1F YES Do you feel that you have it 24 To a very great degree To a significant 7 To a minor degree 24 Have you ever had a course in sociological theory or other social theory where Marx Weber Durkheim Parsons Compte and similar theorists were discussed for at least two weeks NO YES 25 Do you know about the Sociological Department s program Concentration in Analysis and Research No A little Yes fully Part IV Personal Behavior Attitudes and Beliefs Questions 26 How happy are you Very Happy Moderately Happy Unhappy 27 On a seven point scale where would you place your beliefs regarding Abortion Very Pro Completely Very Pro Choice Neutral Life 7 6 5 4 3 2 l 28 Which label best describe your current political beliefsattitudes Very Conservative Very LeftistRadical lLiddle of RoadModerate Moderately Conservative Moderate LeftLiberal Libertarian Not Political at All Other 29 What is your opinion of President Reagan s Policies Do you Strongly Moderately Moderately Strongly No Opinion or Approve Approve Disapprove Disapprove Don t Know 30 Which statement best describes your opinion on the Nuclear Freeze Campaign 1 support President Reagan and oppose any campaign that would weaken our defenses I am willing to permit our defenses to weaken relative to the Soviets and support the extremist AntiNuclear campaign 31 Do you do coke Yes No 32 Most experts note that young adults begin having sexual intercourse between the ages of 15 and 17 without any ill effects and recommend no change At what age do you think it is appropriate for young adults to begin to engage in sexual intercourse Years 33 How frequently do you attend religious services More than once a week About once a week Less than once a week L 4 Do you support the murder of unborn children by Abortionists Yes No Undecided L Lquot Check the statement that expresses your level of comfort with homosexuals I feel uncomfortable around homosexuals I would not enter a gay bar and would feel better if fewer homosexuals were on Madison s sidewal I feel slightly uncomfortable around homosexuals While I tolerate them as a minority group I would not have a close personal friend who was a homosexual I am not bothered by homosexuals If I was at a party where there were homosexual couples or if I had a known homosexual as a coworker it would not disturb me I am completely comfortable around homosexuals I would like to have some gay friends and would not be upset if I was assigned a homosexual as a roommate in the dorms L O What is your opinion about the civil war in Ethiopia Support Ethiopia Support Rebels L l Which statement describe your attitudes regarding feminism I strongly support feminism and oppose all sexism l moderately support feminist values I don t care one way or the other l moderately oppose feminism I strongly oppose feminism and support traditional values Do you think most politicians are dishonest or are they like other people L so Yes No 39 Don t you agree with our President and think that we should increase military aid to governments that fight communists and terrorists Yes No 4 O State whether you Agree Or Disagree With the following statement I don t think that it is an unwise policy to continue sending food aid to Poland 4 Do you think that it is wise to cut spending on social programs and increase military spending Yes No 42 Identify which of the following six occupational groups you would choose to entry if you had to choose among them based on work conditions tasks involved and all forms of reward involved Writer or Artist Forest Ranger or Sailor Top manager or Corp Executive Teacher or Counselor Engineer or Chemist Leader of a Movement for Social Change 43 Eleven career or life goals are listed below choose the three which are most important to you Rank your choices 1 2 an Help others especially the less advantaged Make a large income and accumulate wealth Achieve a position of importance and power in a large organization Care for my family and those close to me Get a secure job without fear of unemployment or layoffs Live in a place where I will be happy and content Make an important contribution to humankind Be independent and control my life without having a boss Become wellknown and popular among a large number of people Advance quickly and become a success in my field Have a chance to be creative and explore new ways of doing things 44 Do you favor the legalization of pot and similar harmful drugs that young people abuse and increase drug addiction Yes No 45 This last question is about friends of yours think of your 45 closest friends when answering it Check the space that best represents how your friends think or the type of friends you have MY FRIENDS ARE Conservative Liberal Athletic Not Athletic Noisy Quiet Studious Not very studious Neat in Dress Sloppy in dress Appearance Appearance Like Classical Like Rock Music Music END OF QUESTIONNAIRE THANK YOU RolePlay Exercise 1 0 Groups of three one Interviewer one respondent one observer 0 Practice asking questions about some topic Eg Plans after graduating Hobbieslrecreation Ideas ora dream vacation Opinions about a current political issue 0 Each group member prepares a different main question and follow ups Roleplay Exercise 2 Interviewer role Practice controlling non verbal behavior responses e c Respondent role answer questions candidly Observer role Keep time interviews should last about 3 minutes Observe behavior of interviewer Facilitate discussion afterward RolePlay Exercise 3 Discussion questions 1 How did the interviewer feel about how it went Easy Dif cult Mistakes Would you do anything differently 2 How did the respondent feel about how it p wen 3 What were the observer39s impressions of the interview Survey Instrumentation Soc 357 Fall 2006 Instrumentation The kinds of questions you ask u The overall structure of the questionnaire m Different kinds of questions will lead to different kinds of answers Openended amp Closedended Questions Openended no limits on answers Closedended a nite set of answers to choose from 7 Eg What did you like most about the University ofWisconsinMadisonquot VS Which ofthe following things did you like most about UW 39son the social life the academic program the athle ics program the location the political orientationquot Yield Different Answers Schuman et al compared answers to open ended and closedended questions about important national problems Each question got a very different distribution of answers The researchers concluded that openended questions were a more reliable gauge of the pulse of the nation Direct amp Indirect Questions 0 Direct questions ask information that is directly connected to vrlnat the researcher is interested in nding out 7 E Q What is your malor7 How many times a week do you eXefClSe 7 0 Indirect Questions the link between the researcher s objectives ampthe question is less clear sed to address serisiriye ropics Different Response Formats 0 YesNoDon t know 0 Likert response scale 0 711 point numbered sca es 7 More reliaole ampyalid than smaller number of categorles o Adjectives for each possible answer 7 E 9 Strong Democrat Weak Democrat Strong Republican Weak Republican Other MultipleChoice Likert Format Strongly SWle Disagree Agree Nwtml Disagree Agree 1 2 3 4 5 Never Rarely Sometime Often Always 1 2 393 4 s Composite measures 0 Used for concepts that are complex amp hard to measure 0 Ask multiple questions for each concept 0 Use answers to construct an Index a sum or average of he scores for each question related to one concept 0 Or to construct a Scale sum of answers to a series of questions that vary along an underlying continuum Example of an Index Trust in government Sivanglv agree Sivanglv Dl gvee F39ulltlcal ufflclals eeh t eare abuut peuple llke me 2 3 4 5 in general l belleve pelltleal em lals are hehest l 2 3 4 5 lh general l trust pelltleal ufflclals l 2 3 4 5 tn t the rlghtthlhg Overall l thlrlk that pelltleal ufflclals are Werhhg furthe bene t l 2 3 4 5 at all Index Score wouldrbe gaunt or averageoral thesmswers Example of a Guttman scale Political Action Response Pattern v v Y Y Signed a petition Participated in a boycott Y Y Y Y Attended a lawful demonstration Y Y Y N Joined an unof cial strike or v v N N demonstration Occupied a building orfactory Y N N N Scale Score 4 3 2 l u Structuring Questionnaires Opening questions Sensitive amp Routine Questions Transitions from topic to topic Providing a frame of reference Question sequences mewm 1 Opening Questions The goal get your informant warmed up amp comfortable wi t e role Openended questions allow expression of views of success comfort WI h But must be consistent with what has been said about the purpose of the interview Easyto answer questions generate feelings t role 2 Sensitive amp Routine Questions Avoid placing either type at the beginning need to build trust Boring questions often at the end Sensitive questions often in the middle Keep ow of questions logical 3 Topic Transitions o Transitions are a part ofnonnal social interaction 7 Now l d like to ask you afew questions abou x 0 Early topics should be easy to answer 0 Later topics should ow logically from the earlier ones 0 Be careful of questionorder effects the possibility that earlier questions affect answers to later uestions e E g Asking about a belief in srnoking amp lung cancer tnen asking about costs amp benefits of srnoking 4 Providing a Frame of Reference Important to understand what is the respondent s frame of reference Find this out through 7 question sequencing 7 providing context 7 providing time 39ame 7 providing facts 7 providing vocabulary 4 Frame of reference Examples There are cases in which a man and aWOman hyetogetheh set up housekeeping have children take care ot them but do not get a marriage license Do you think children otthese common iaw marriages shouid haye ihhehtahce rights AS you may have heard if you Were at the last School board o a will hayeto iay ott about 20 ot the teachers in Septem er Given this Situation do you think people shouid vote for the tax Question Sequences 0 Moving from general to speci c questions funnel sequence or Viceversa reverse funnel sequence 7 E 9 What do you think about Social iite at uvv Madisom to Do you beiieye there l5 a problem With bihgedhhkihg arhohg uhd rgraduates 0 Using reason analysis breaking dovm a decision process to ask about different stages many whys Instead of one Vin quot e g vv y UCLA gtvvheh didyou Slanlhinking about going to college vvhat colleges didyou CoriSider whatwere some of the things you iiilted especiaiiy about UCLA Response Biases Social desirability of some answers 7 olutions question placement wording use of indirect questions Tendency to be agreeable e Solu ion include both positive amp negative phrasings Don t rememberfalse memo 7 Solutions Shorten reference period more time to answer provide context Steps in Writing a Questionnaire 1 Identify the objective of your study 7 Possibly do focusgroups literature review 2 SketchOutline the topics you want to cover in your questionnaire 7 Think about placement overall ow 3 Fill in the items 7 Think about different types of ques ions wording 4 Pretest amp Revise Slide 1 Slide 2 Slide 3 Measurement amp Validity in Your Questionnaire Assignment Reliability and Validity Ruliallle Valm Nul Vand Nut Rnlialxlu Neillmr Rolialtlu Nor v Ball Ruliabln nlill And Valid cwmmm VIIam MK Trodim All Rigs Ramed Two general approaches to assessing construct validity Constluct Validity issue 15 do you have a good operauonalrzatron of your construct the thing you are trying to measure7 want to measure or operauonallze7 I want to know how In love people are e ineuieeantentafuierneasure duesltmakesense7 7 Study its relaaanta uLhEr variables Slide 4 Slide 5 Slide 6 Construct Validity Content This is essentially definitional does the content ofthe measures match what you have in mind by the concept or construe 7 cme validityquot you look atthe measure and it makes sense quot The weakest criterion Cnntent validityquot you more systematically aiamlne or inventory the aspects ofthe construct and determine whether you have captured them in your measures You may ask others to assess whetheryourmeasures seem reasonable to them Construct Validity by Criterion Predictive it correlates in the expected way with ething it ought to be able to predict E g love ought to predict marriage or gazmg7 should be able to distinguish between e g between those dating each other and those not correlate with each other this is whatwe are doing in the survey exercise Divergent able to distinguish measures ofthls construct stru oving vs from related but ditrerent con cts eg 1 liking or knowledge from testrtaklng ability Convergent Validity and Reliability Convergentvalidity and reliability merge as concepts when we look atthe correlations among different measures ofthe same concept Key idea is that lfdlfferent operationalizations measures aremeasurin th sa econcet construct they should be positively correlated w eachother People who appear ln lovequot on one item should ar u in lovequot on another item And conversely Slide 7 Slide 8 Slide 9 Steps in Data Analysis 1 Check for Errors First check for errors Look a1 list output Check data back against the surveys See example listing 1 in doc le Slide 10 Slide 11 Slide 12 2 Check Frequencies 0ut ofrange values are errors that needto be xed Look for problems of low van39ability See examples offrequencies in doc le Low Variability is Always a rob e vaarlables do notyary they cannot have statistical association with oLhervanables Thls is always a problem in statistical analysis Lowrvanablllty i ems may indicate blased questions Lowrvanablllty items may indicate thatyou unfortunately got a biased sample Lowrvanablllty items may indicate thatpeople Just do noty much on that o inion orbehavlor e g oxygenrbr rig b important but cannotbe eathi May e studied using statistical associations 3 Examine Reliability Analysis The reliability analysis tells us whether your closedended dependent variable items appearto be measun39ng the same concept Slide 13 TrueScore Tlmy Sin i ed mmaryqnsdmisdetoaomumamcf se True Random s cc A Error PM cwgtmm VIIam MK TM Slide 14 InterItem Correlations are an Estimate of Proportion of Variability Due to True Score Rather Than Error Slide 15 Ways of Assessing Reliability from InterItem Correlations Could assess average mlemtem correlath Could assess average ltemrtoml correlathn Could assess sphthalfconelauoh Alp a IS the equlvalmt ofall posslble sphthalf correlatlons although nothow ll 15 computed Following graphlcs are from Wllllam M K Trochlm s ohrhhe methods book Copyrlght ZOOZAllng1ts Reserved http Lrochlm human Cornell edukb Slide 16 Slide 17 Slide 18 Average InterItem Correlation Ham quot m I2 99 m n I 91 92 um um l 1 m 9 351m m ins w m Ln maqu L m 91 as 37 5101 H va 39 90 cwmmmymnamMKTm 394 394 39x k NJquot 95 1 W cwgtmm VIIam MK TM SplitHalf Correlations E cwgtmm VIIam MK TM Slide 19 Gubadq s Alfiarsmscalyawayof sunrranzrgthe usarru gall en39s F r SH SH SH El SH as o 85 SHs w SH as Slide 20 Results of Reliability Analysis See oramples or doc file r 01 all posltwe moderate to strong correlatlons alpha pretty hlgh gt 7 or 50 OK som ye conelatlohs ho negatwe oorrelatrohs alpha hot awful gt 4 or 5 anyway ltems notnotlceably dlfferent from each other PROBLEMS ANDOR all 1 not scalequot some negatwe correlationsltr1 orrelatlons are hearzero The ltems do Slide 21 Solutions to Problems 1n Reliability Get Help From Me Most correlatlons posltlve Wth hlgh ltemrtotal conelatrohs but a few rtems haye negatwe correlatlons with the others lndlcates you need to drop a few ltems from the lndex usually lndlcates that some need to be reverserscored often lndlcates codlng errors especlall t am members reverse scoredrdlfferently Slide 22 Slide 23 Slide 24 If the Items Do Not Scale ow mixed positive and negative correlations m indicate that the items do not scale quot That is th e not measuring the same concept the one question thatbest th g a E lt The openrmded question can sometimes e helpful in deciding what to do Need Original Questionnaires 1 there are problems with your reliability we will probably need to look at your original questionnaires and how they were coded 39s is the my to detect coding error and figure out what patterns ofresponse may it e Forming the Index We are not studying statistical reliability in depth Justuslng it as a tool to pick outDv questions tha have the highest possible lnLa dLem correlations Index 2 itemi 4 2m The theo is that the random errors in the items will cancel out so the sum ofthe items the lndai will have LESS ERROR lte at a time Your index is your measure ofyour dependent p varlableconce t The love scalequot IS the measure of being in love Slide 25 Slide 26 Slide 27 Index Dependent Variable We told the computerto compute your index as the sum ofyour closedended questions You can check this by adding up the scores for one person and seeing ifyou get e same index We should have documented this better Convergent Validity with Open Ended Question You can check your openended question and the index against each other You MUST know What the MEANING of the different categories ofthe openended question is to do the interpretation See example in doc le ofone that works Testing Your Hypothesis Difference ofmeans dependent variable index by categories ofindependent variables Correlations between dependent variable index and intervallevel independent Variables Slide 28 Slide 29 Slide 30 WriteUp Questionnaire Refer to Assignment This is just a snoit outline ofpoints to emphasize Style Notes 1 Refer to variables by consistent names to computervariable names E g abortion is mur e ABMUR 2 Tables must be labeled 7 Variable names sufficiently aiplained e i i r numb ers Beg inning TitlePage 39 Abstract Introduction AS BEFORE Slide 32 Slide 33 Slide 31 Sampling Exp1ain how you got the subjects Don t worry about convenience sample goo Do consider whether you think yo got a d mixture ofpeople for the topic you are studying Measures independentvariabies the are the questions you used Briefevaluation ofwhether er were any problems Dependentvariabie closedrmded questions 7 Discuss development 7 De nehigh end ufscale 7 Ex 1 en rsescunng withtwu examples Dependent variable openrended question give details on coding how y u grouped the a w explanations ofhow you decided who went into each group 5 Validity of Dependent Variable VanabiIity either discuss prob1ems or say why OK and give ex 39 39 the worst vari i1ity Briefsummary ofre1iabi1ity index construction Open by 1e ofvanable With recopy tab1e with LABELS for the open categories OK to rearrange them Discuss whether they are consistent Slide 34 Results Univariate Brief discussion offrequencies Bivariate Test ofhypothesis with di erence ofmeans for index or ion PREPARE TAB Other bivariale results PREPARE TABLES Discussion Slide 1 Slide 2 Slide 3 Sociology 357 Lecture Notes 1 Basic Methodological Concepts Methods of Sociological Inquin Getting Sta ed Why bother to learn how to do research7 7 Answer quesnuns irnpurtant tu us uns er ufuthers research true urfalse e Truthelsehuud eanhe deterrnrnedhy ubservatlun Values 7 atuughttuhewhatisguudurhad e Disanguished 39nm ernpineal statements 7 In uence seleeaun ufresearehtupies Nonscientific Approaches to Knowledg e Unsupported assertions e no evidence at all Appeals to authority alien e Unconcretized abstractions e g aggressive 7 specific behaviors unclear 7 Errors of observation Slide 4 Slide 5 Slide 6 Errors of Observation Inaccurate observation Simply Wrong about what you think you saw Overgeneralization Correct about what you saw but apply it too broadly Selective observation Only notice what supports your View Scientific Approach to Knowledge Empiricisrn take sensory input from Lheworld Objectivity differentpecpie can observe the same inputs and see the same thin 7 Reliability ubseryauuns eanbe eunsistentiy rnade uyer urne and by differmt peuple ab nut ubservanun e Vah ity theoreaeaieuneeptuai interp etatiun is euneet about link between ubseryaaun and ufubseryatiuns theury Controlled observation to eliminate sources of aror 1 Observation Knowledge as tentative subject to further refinement Scientific Attitude Knowing better than not knowing ruth to selfdelus Always ten the truth about your research atter What Respect the evidence prefer unp1easant ion nom Recognizes the boundaries of expertise Slide 7 Slide 8 Slide 9 Studying People Freewlllvs soelalpallems 7 Sunal cunsh mnts un ehnl s money laws 7 Sunal effects un nhmcesleadpeuple tn wanna hehave ln eenan ways ProbablllsLICLhmkmg 7 Reenonze hnlh the general tendency and the Venetian aruundthattmden Understanding venslehen Vs prediction explanallon The Scientific Process Emplneal Gma39allzanuns Predlmuns hypalheses Theory Not ldle speculation but coherent accouan of work Theory 15 part ofsclence making sense ofwhat Lhe ollsllnol observations mean how they are related re oononeuzeol theoretical elemenls ons lo emplnoal phenomena Explanauon and prediction as equwalenl Abstractlons a hav 6 connectl Slide 10 Slide 11 Goals of Scientific Research Descrrptron Accurate desenptron ofwhat ls actually happenlng ls the bedrock ofsclence ur dlfferenees between subjects Explanatron and predretron Provlde a theoretreal account for obseryatrons as part ofa larger g o how the world works and predrrt L 5 Understandrng ofpeople why they do what they do Ofsoclal systems how they work and why they work thatway Research Steps Deflne a problem oflnterest Idmtlfy emplrlcal questrons wrthrn thatproblem Useereate a general theoretreal framework for understandrng the problem Concreu39ze the concepts in the theory 7 operau39onalize the Valiables vgt E s E dat Draw concluslons Slide 1 Asking Questions A few pointers Sllde 2 MultipleChelce Likert Format Strungly Strungly Disagree Ages Neutral Disagree Ages 1 2 3 4 5 Never Rarely Sumeumes O en Always 1 2 3 4 S Sllde 3 Match response to item 39 Frequency NeverAll the time 39 Likert Scaling DisagreeAgree 39 Quality PoorExcellent 39 Service Not WellExtremely Well 3 Slide 4 Slide 5 Slide 6 Use Simple Sentences 39 No double negatives 39 Eliminate vagueness poorly de ned S 39 Avoid objectionable1rrelevant questions Avoid Double Negatives Avoid Vague or Ambiguous Terms Slide 7 Avoid Objectionable or Irrelevant Slide 8 Avoid Doublebarrel Questions Slide 9 Balance questionsresponses Slide 10 Reverse Score to Reduce Response Bias Slide 11 Exhaustive amp mutually exclusive cate OIies Slide 12 Even vs Odd catego es Measurement amp Validity in Your Questionnaire Assignment Reliability and Validity l 371 l l39 l quot llt393g3939l l quot391 i i J U N V Reliille Mi Neither Reli ll Bmh Rllilhle ullhld nt eliahla analid AndValid Copyright 2002 William MK Trochim All Rights Reserved 39 human mmell edlllzh Construct Validity I Issue is do you have a good operationalization of your construct the thing you are trying to measure I First it is necessary to de ne the construct What is it you Want to measure or operationalize I Want to know how in loVe people are I Two general approaches to assessing construct Validity 7 Examine the content ofthe measure does itmake sense 7 Study its relation to other Variables Construct Validity Content This is essentially de nitional does the content of the measures match what you have in mind by the concept or construct Face validity you look at the measure and it m e s The weakest criterion Content validity you more systematically examine or inventory the aspects of the construct and determine whether you have captured them in your measures You may ask others to assess whether your measures seem reasonable to them I Conver Construct Validity by Criterion Predictive it correlates in the expected way with something it ought to be able to predict Eg love ought to predict marriage or gazing Concurrent able to distinguish between groups it should be able to distinguish between eg between those dating each other and those not ent different measures ofthe same construct correlate with each other this is what we are doing in the survey exercise Divergent able to distinguish measures ofthis construct from related but different constructs eg loving vs liking or knowledge from testtaking ability Convergent Validity and Reliability I Convergent validity and reliability merge as concepts when we look at the correlations among different measures ofthe same concept Key idea is that if different opemtionalizations measures are meas 39 concep construct they should be positively correlated with each 0 er I People who appear in love on one item d appear in love on another item And conversely Steps in Data Analysis 1 Check for Errors 0 First check for errors Look at list output Check data back against the surveys See example listing 1 in doc file mi nf nu of Data 2 Check Frequencies 0 Out of range values are errors that need to be fixed 0 Look for problems of low variability 0 See examples of frequencies in doc file Low Variability is Always a Problem If variables do not vary they cannot have statistical association With other variables This is always a problem in statistical analysis Lowvariability items may indicate biased questions Lowvariability items may indicate that you unfortunately got a biased sample Lowvariability items may indicate that people just do not vary much on that opinion or behavior eg oxygenbreathing May be important but cannot be studied using statistical associations 3 Examine Reliability Analysis The reliability analysis tells us whether your closedended dependent variable items appear to be measuring e same concept TrueScore Theory Simpli ed Answer to any question is due to a combination of true score plus error Observed True Random Score Ability Error Copyright 2002 William MK Trochim All Rights Reserved InterItem Correlations are an Estimate of Proportion of Variability Due to True Score Rather Than Error Ways of Assessing Reliability from InterItem Correlations I Could assess avemge interitem correlation I Could assess avemge itemtotal correlation I Could assess splithalfcorrelation Alpha is the equivalent of all possible splithalf correlations although not how it is computed Following graphics are from William MK r chim s online methods book Copyright 2002 All Righw Reserved I httptrochim humancornelledukb Average InterItem Comla on Copyright 2002 William MK Trochim All Rights Reserved llI39l 52 IIll Copyright 2002 William MK Trochim All Rights Reserved Split Half Correlations Copyright 2002 William MK Trochim All Rights Reserved Cronbach s is basically a way of summarizing the correlations among all items Cronbach s alpha a Ir rr n x aH a7 SH a5 5H 91 aH a3 a 85 sea as ah a5 CumHUM EIEI HHam M K Tmrmm H R E H M RE ENEd Results of Reliability Analysis I See examples in doc le I GOOD all positive moderate to strong correlations alpha pretty high gt 7 or so I OK some positive correlations no negative correlations alpha not awful gt 4 or 5 anyway items not noticeably different from each other I PROBLEMS some negative correlations ltl ANDOR all correlations are near zero The items do not scale Solutions to Problems in Reliability Get Help From Me I Most correlations positive with high itemtotal correlations but a few items have ne a 39ve correlations with the others indicates you need to drop a few items from the index I Mixture of strong positive and negative correlations usually indicates that some need to be reversescored I All low mixed positive and negative correlations o en indicates coding errors especially team members reverse scoreddifferently If the Items Do Not Scale I Low mixed positive and negative correlations may indicate that the items do not scale That is they are not measuring the same concept I In this case you pick the one question that best measures What you had in mind and an that correlate with it for your dependent Variable I The openended question can sometimes be helpful in deciding What to do Need Original Questionnaires I If there are problems with your reliability we will probably need to look at your original questionnaires and how they were coded I This is the way to detect coding errors and figure out what patterns of response may mean Forming the Index We are not studying statistical reliability in depth just using it as a tool to pick out DV questions that have the highest possible interitem correlations Index item1 itemZ item3 item4 etc I The theory is that the random errors in the items will cancel ouL so the sum of the items the index Will have LESS ERROR of measurement than just one item at a time Your index IS your measure of your dependent Variable concept The love scale IS the measure of being in oVe Index Dependent Variable 0 We told the computer to compute your index as the sum of your closedended questions 0 You can check this by adding up the scores for one person and seeing if you get the same index 0 We should have documented this better Convergent Validity with Open Ended Question 0 You can check your openended question and the index against each other 0 You MUST know what the MEANING of the different categories of the openended question is to do the interpretation 0 See example in doc file of one that works Testing Your Hypothesis 0 Difference of means dependent variable index by categories of independent variables 0 Correlations between dependent variable index and intervallevel independent variables WriteUp Questionnaire Refer to Assignment This is just a short outline of points to emp asize Style Notes 0 1 Refer to variables by consistent names that reflect question content and are linked to computer variable names Eg abortion is murder ABM UR 2 Tables must be labeled 7 Variable names suf ciently explained 7 Content of openended categories listed next to numbers Beginning Title Page Abstract Introduction AS BEFORE Sampling I Explain how you got the subjects I Don t worry about convenience sample I Do consider whether you think you got a good mixture of people for the topic you are studying Measures I Independent variables these are the questions you used Brief evaluation of Whether there were any problems I Dependent variable closedended questions 7 Discuss development 7 De ne high end ofscale 7 Explain reverse scon ng with two examples I Dependent variable openended question give details on coding how you grouped the answers With explanations of how you decided Who Went into each group Validity of Dependent Variable I Variability either discuss problems or say why OK and give example of variable with the worst variability I Brief summary of reliability index construction I Open by index recopy table with LABELS for the open categories OK to rearrange them Discuss whether they are consistent Results Univariate Brief discussion of frequencies Bivariate Test of hypothesis with difference of means for index or correlation PREPARE TABLES Other bivariate results PREPARE TABLES Discussion Slide 1 Sampling Slide 2 Why Sample Why not study everyone Debate about Census vs sampling Slide 3 Problems in Sampling What problems do you know about What issues are you awe op What questions do you have Slide 4 Slide 5 Slide 6 Target pupulauun pupulauun uflnterest pupulauuntu Which Generalizatauns Are Made De nedLlsted hy Sampling Frame Target Sample Rasponsa not Sample The people actually studied Who do you wall a The Theoretical generalize to population Wllal papulalmn can me slimy you get access in Papulxlnn Haw can ynu get p Tne sampling access Io them l i Frame wno ls In your sway l The Sanple apyng1 200 llllllam MK TM Sampling Process UnlLS of Analysls people List or Procedure Sampling Frame List ufTarget Sample Key Ideas e sampling frame actual population Understand crucial role ofthe frame Distinction between the popul interest and the actual population de ned b Generalizations can be made only to the alien of sampling Slide 7 Slide 8 Slide 9 Sampling Frame The list or procedure defmmg Lhe POPULATION rom which the sample Wm be drawn Distinguish sampling fmme from sample 7 Randam 34g hahhg Essmtial for probability samphng but can be defined fornonprobability samphhg Types of Samples Probability Samples A probability sample is one in which each element ofthe population has aknown nonrzero probability ofselection Not aprobability sample ofsome elements ofpopulanon L 39 39 39 Not aprobability sample ifprobabilities of selection are not known Slide 10 Slide 11 Slide 12 Probability Sampling Cannot guarantee representativeness on all traits of interest A sampling plan with known statistical properties permits statements like quotThe probability is 99 that the true population correlation falls between 46 and 56 Sampling Frame is Crucial in Probability Sampling lfthe sampling frame is a poor fit to the population ofinterest random sampling from that lem The sampling framers nonrandome chos n em ts not in the sampling frame have zero probability ofseleetion 39 Generalizations can be made ONLY to the actual population defined by the sampling frame Types of Probability Samples Slide 13 Slide 14 Slide 15 Sin e Rancan Slide 16 Slide 17 39Hsmemmas mnmiftl39g stis inqu m it e sd inp citsim m mifh EStisgmq d Slide 18 Rush mf if Pm utymmehgm the mism I hm ezevay4th 818121115186 m Slide 19 Slide 20 Slide 21 Random Cluster Sampling l wedly this is a form ofrandom sampling divided rnlo groups usually r orgamzauonal Done co Populahon ls geogaphlc 0 Some othe groups are randomly chosen in pure cluster samphng whole cluster is sampled In simple multistage clusler there is random sampling within each randomly chosen clusler RandornOust 39 Sangiing Z mmdsdmdedmgwps mdhmmmmdy seleoed Fagvensangeslzeaclma ehsrrneeuumanasinge ranh39nsarqie mmdchmrgnay pannlagetsa39qie Eiuissrmlletif ech elsale Whom Slide 22 Slide 23 Slide 24 Stratification vs Clustering Stratificatiun Divide pupulanunmtu gmups 6433mm sum eae e er sexesra es ages Samplerandumly sum each guup Less enurcumpzredtu slmplerandnm luster Divide ing pupulauunmtu cumpzrable guups seheeis Clues Randumly sample seme er the gmups Mare murcumpzredtu slmplerandum stxan canunmfm manun befure sampling unly eeme areas er er n5 gamzanu Stratified Cluster Sampling Slide 25 Slide 26 Slide 27 Multistage Probability Samples 71 Multistage Probability Samples 72 ll ra l The Problem of NonResponse 1 You can randomly plck elements from sampllng frame and use them to randomly select people Butyou cannot make people resp Norrresporlse destroys the generalrzeabrlrty othe samp e wllllng to respondto surve s 39 Ifresponse ls 90 or so not so bad But lflt ls 50 ths ls a senous problem Slide 28 Slide 29 Slide 30 The Problem ofNonResponse 2 Multiple callbacks are essenual fortrylng lo reduce nonrespens e blas Samples Wlthout callbacks have hlgh blas cannoL really be consldenedimndo m samples Response39rales have been falllng It lsvery dlfflcult lo get above a 60 response lame n mate the effect oflhe39e mrby getting as much mfom39aallon as posslble aboutthe predlcwrs ofnonrresponse Slide 31 Slide 32 Slide 33 Quota Vs Stratified Sampling m ed sampling Withuut Macks may nupmwcdce be much iff ml gamma gtsamphngk Butyou should knbwthe dilferenceffdrthe fest Slide 34 Sample Size Helerogenelty need larger sample to study more dwerse populauon Deslred preclslon need larger sample to get smaller error Sampllng deslgn smaller lfslraufled larger If clusler Natureofanalysl s complex multlvanate statlstlcs need larger samples Accuracy of sample depends upon sample slze not rauo ofsample lo populatlon Slide 35 Sampling in Practice O en a nonrandom selecuon ofbaslc sampllng frame clty organlzatlon elc gt Fltbetween sampllng frame and research goals must be evaluated Sampllng frame as a Concept 15 relevantto all klnds ofreseamh lncludlng nonprobablllty Nonprobablllty sampllng means you cannot generallzebeyond the sam le Probablllty sampllng means you can generallze Lo opulauon Asking Questions Soc 357 Fall 2006 Methods of research Surveys Interviews Gathering Data questions items Openended amp Closedended Questions I Openended no limits on answers I Closedended a nite set of answers to choose from 7 Eg What did you like quot 39 quot f WisconsinMadisonquot VS Which ofthe following things did you like most about UW Madison the social life the academic program the athletics program the location the political orientationquot Types of Questions on Surveys I Social background information I Reports of past behavior I Attitudes beliefs values I Behavior intentions I Sensitive questions Types of Questions in Interviews I Anything but especially questions that elicit descriptions attitudes stories in other words questions that try to get the respondent to describe things in his or her own words rather than choosing from a set of answers given by the researcher Conducting Surveys o Surveys followa strict protocol to minimize their impact on the respondent 7 Reading questions exactly as Written 7 Use oredesignateo oiooes peneended questions Witnout comment 7 Do not give personal intonnation express opinions or give feedback 0 Eg Lavin amp Maynard Laughter in Survey Interviewsquot Conducting Interviews Interviewers maintain openness willingness to listen nonjudgemental attitude put ego aside Establish a communicative settin e Askin egin 7 Displaying recognition amp empathy e Controlling nonverbal behavior eye contact dgeting body position facial expression tone of voice distracting mannerism 9 by gnonthreatening questions especially at he ning Analysis of Data Survey Interviews Preparation Coding ans Coding rs an wers numerically thematically Results Description N Description D Explanation Theory Survey Strengths amp Weaknesses o Strengths 7 Broader range oftopicsthan experiments 7 Ef cient Way to gamer a lot of data 7 Generalizeable rr usrrrg probability sample 0 Weaknesses 7 tisolate causal variables theoretical aSSOClathn between variables rs always imputed by the researcher 7 Bad question Wording undermines reliability amp Validity of questions pretesting is key Interview Strengths amp Weaknesses Strengths 7 Data is rich helps us understand why we see associations 7 You can ask for clari cation connections 7 Less costly than surveys Weaknesses 7 Results not generalizable 7 Results aren t compact Comparing Ransford AND Ewick ancl Silbey For each article what were the 7 Units of analysis 7 Sample 7 Variables 7 Operationalization ofvariables 7 Re Its 7 Strengths ampweaknesses Interaction Effects I When two variables together affect a dependent variable differently than either one of them would on its own 7 Eg Alcohol amp sleeping pills either one will make you sleepy but taken together they interact and knock you out 7 Eg Ransford s study Poweriessness amp Dissatisfaction taken together they increase the likelihood ofvio ence by more than ifyou added up the individual effects of each Spuriousness I Can some other variable account for the statistical association you see I To check we hold constant other possible explanatory variables we look within each category of the extra variable to see if the association we originally observed still holds 7 Eg Smoking amp Lung cancer amp Social class 7 Ransford held Neighborhood amp Education constant Ransford Vs Ewick amp Silbey I What background assumptions is each article making I What do we learn about the phenomenon of resistance from each article I Which finding do you think is more interesting Why Types of Probability Samples Simple Random Sampling Each element in the population has an equal probability of selection AND each combination of elements has an equal probability of selection Names drawn out of a hat Random numbers to select elements from an ordered list Strati ed Random Sampling1 Divide population into groups that differ in important ways Basis for grouping must be known before sampling Select random sample from Within each group Strati ed Random Sampling2 For a given sample size reduces error compared to simple random sampling IF the groups are different from each other Tradeoff between the cost of doing the stratification and smaller sample size needed 0 for same error Probabilities of selection may be different for different groups as long as they are known Oversampling small groups improves inter group comparisons Systematic Random Sampling1 Each element has an equal probability of selection but combinations of elements have different probabilities Population size N desired sample size n sampling interval kNn Randomly select a number j between 1 and k sample element j and then every kth element thereafter jk j2k etc Example N64 n8 k6488 Random j3 Systematic Random Sampling2 9 Has same error rate as simple random sample if the list is in random or haphazard order Provides the bene ts of implicit strati cation if the list is grouped Systematic Random Sampling3 Runs the risk of error if periodicity in the list matches the sampling interval This is rare In this example every 4 11 element is red and red never gets sampled If j had been 4 or 8 ONLY reds would be sampled Random Cluster Sampling 1 Done correctly this is a form of random sampling Population is divided into groups usually geographic or organizational Some of the groups are randomly chosen In pure cluster sampling Whole cluster is sampled In simple multistage cluster there is random sampling within each randomly chosen cluster Random Cluster Sampling 2 5 a Population is divided into groups Some of the groups are randomly selected For given sample size a cluster sample has more error than a simple random sample Cost savings of clustering may permit larger sample Error is smaller if the clusters are similar to each other Random Cluster Samplng 3 Cluster sampling has very high error if the clusters are different from each other Cluster sampling is NOT desirable if the clusters are different It IS random sampling you randomly choose the clusters But you will tend to omit some kinds of subjects Strati cation vs Clustering Strati cation Clustering Divide population into Divide population into groups different from each comparable groups other sexes races ages schools cities Sample randomly from Randomly sample some of each group the groups Less error compared to More error compared to simple random simple random More expensive to obtain Reduces costs to sample strati cation information only some areas or before sampling organizations Strati ed Cluster Sampling Reduce the error in cluster sampling by creating strata ta 3 of clusters Sample one cluster from each stratum The costsavings of clustering with the error reduction of strati cation Strati ed Cluster Sampling Combines elements of strati cation and clustering First you de ne the clusters Then you group the clusters into strata of clusters putting similar clusters together in a stratum Then you randomly pick one or more cluster from each of the strata of clusters Then you sample the subjects Within the sampled clusters either all the subjects or a simple random sample of them Example of List of Data Page 1D JEWCULT JEWFRNDS SELFJEW CHEAP WSNOBS ECONWELL REPPERSC DISCTODY OPENEND 10100 400 3 00 300 200 3 00 3 00 400 100 100 10200 400 5 00 200 400 3 00 3 00 400 300 400 10300 500 4 00 200 300 3 00 3 00 400 200 300 10400 500 5 00 300 300 3 00 5 00 400 200 400 10500 400 5 00 100 300 3 00 4 00 400 300 100 10600 500 5 00 300 300 2 00 4 00 500 400 400 10700 400 5 00 300 400 4 00 3 00 400 200 300 10800 200 3 00 100 400 4 00 2 00 200 200 10900 300 2 00 100 300 4 00 4 00 100 200 300 11000 100 4 00 100 300 3 00 4 00 400 300 300 20100 200 3 00 100 200 3 00 4 00 300 300 400 20200 200 3 00 100 400 3 00 4 00 200 200 400 20300 200 2 00 100 500 5 00 1 00 500 500 300 20400 100 1 00 100 300 3 00 3 00 400 400 200 20500 100 1 00 100 300 4 00 3 00 400 200 20600 200 3 00 100 500 4 00 1 00 500 500 300 20700 500 4 00 300 300 3 00 4 00 300 200 4 00 20800 200 2 00 100 500 5 00 1 00 400 200 20900 200 5 00 100 300 2 00 4 00 400 400 300 21000 500 5 00 300 400 5 00 4 00 400 300 400 30100 200 2 00 100 500 5 00 1 00 400 100 300 30200 200 2 00 100 400 4 00 4 00 350 400 200 30300 200 2 00 100 500 5 00 2 00 400 400 300 30400 200 2 00 100 500 5 00 2 00 500 300 200 30500 400 3 00 100 300 4 00 4 00 400 200 300 30600 200 2 00 100 200 2 00 4 00 500 200 400 30700 400 5 00 300 400 4 00 2 00 400 200 400 30800 500 5 00 300 500 4 00 2 00 500 400 300 30900 400 5 00 300 300 3 00 4 00 350 400 400 31000 200 2 00 100 400 3 00 3 00 400 300 400 Example of Frequency for Item with No Problem of Variability JEWFRNDS Valid Example of Frequency Table for relatively lOW variability 85 of cases are in 100 200 300 400 500 Total Frequency 2 9 6 3 10 30 Percent 67 300 200 100 333 1000 Valid Cumulative Percent 67 300 200 100 333 1000 categories 4 amp 5 23 in 4 Possible problem THREAT Valid 100 200 300 400 500 Total Frequency Percent 1 2 3 26 7 9 3 26 51 77 667 179 1000 Percent 67 367 567 667 1000 Valid Cumulative Percent 26 51 77 667 179 1000 Percent 26 77 154 821 1000 Example of Frequency Table with serious variability problem THREAT Valid 100 200 300 400 500 Total Frequency Percent 0 0 0 923 77 1000 Valid Percent 0 0 923 77 1000 Cumulative Percent 0 0 923 1000 Example of Reliability That Looks Good Method 2 covariance matrix will be used for this analysis RELIABILITY ANALYSIS SCALE ALPHA Mean Std Dev Cases 1 NEEDWAR 36410 13667 390 2 MAI 30769 12005 390 3 BLEMISH 31282 10804 390 4 ACE 24872 10227 390 5 THREAT 39231 8393 390 6 SUFFER 30641 12470 390 7 SAFER 28205 9966 390 8 MEDIA 29231 11094 390 9 PRIVACY 29103 12767 390 10 ASHAMED 34359 12523 390 11 NEGOPIN 26667 12425 390 Correlation Matrix NEEDWAR REMAIN BLEMISH PEACE THREAT NEEDWAR 1 0000 7711 10000 BLEMISH 5310 4994 10000 PEACE 6180 6546 3469 10000 THREAT 4341 2672 3594 2594 10000 SUFFER 5312 5240 4235 4185 2060 SAFER 4924 6937 3641 5270 2977 MEDIA 6756 6566 6232 4746 2478 PRIVACY 4034 3738 1707 3569 1776 ASHAMED 7397 7823 5411 6929 3082 NEGOPIN 5476 5645 5227 3383 3533 SUFFER SAFER MEDIA PRIVACY ASHAMED SUFFER 10000 SAFER 5283 10000 MEDIA 6504 5108 10000 PRIVACY 2103 1938 4316 10000 ASHAMED 6136 6337 7446 4778 10000 NEGOPIN 5577 5454 6109 2378 6032 NEGOPIN NEGOPIN 10000 R E L I A B I L I T Y A N A L Y S I S S C A L E A L P H A N of Cases 390 N of Statistics for Mean Variance Std Dev Variables Scale 340769 858492 92655 11 Item total Statistics Scale Scale Corrected Mean Variance Item Squared Alpha if Item if Item Total Multiple if Item Deleted Deleted Correlation Correlation Deleted NEEDWAR 304359 661603 8016 7297 8948 310000 682763 8131 7721 8946 BLEMISH 309487 736420 5953 4834 9062 PEACE 315897 734720 6463 5586 9039 THREAT 301538 793573 3870 3043 9145 SUFFER 310128 708485 6405 5195 9042 SAFER 312564 736299 6564 5968 9036 DIA 311538 698836 7944 7121 8962 PRIVACY 311667 752149 4066 3213 9173 ASHAMED 306410 665388 8685 7856 8910 NEGOPIN 314103 703667 6687 5307 9026 Reliabili 11 items Alpha Standardized item alpha 9099 Weakest items are marked in red but they are still OK Method 2 covariance matrix will be used for this analysis Example of Reliability Suggesting Big Problems R E L I AB I L I T Y A N A L Y S I S S C A L E A L P H A Mean Std Dev Cases 1 ATTRACT 39730 6449 370 2 PARTRES 19459 9412 370 3 PEOPLE 34054 9849 370 4 MYAPP 37568 8946 370 5 OTHERS 43784 5940 370 6 PRIDE 37027 7403 370 7 TLOOK 37027 8454 370 8 TTRY 27297 13049 370 Correlation Matrix ATTRACT PARTRES PEOPLE MYAPP OTHERS PRIDE ILOOK ITRY ATTRACT 10000 the best correlation we may just use these two PARTRES 4094 10000 PEOPLE 1052 0842 10000 MYAPP 1562 1810 0835 10000 OTHERS 1176 0376 O796 1879 10000 PRIDE 1337 2230 0175 0555 1366 10000 ILOOK O661 2651 1515 1350 1570 2339 10000 ITRY 1561 0122 1309 0373 OO77 3746 0259 10000 N of Cases 370 N of Statistics for Mean Variance Std Dev Variables Scale 275946 63033 25106 8 R E L I A B I L I T Y A N A L Y S I S S C A L E A L P H A Item total Statistics Scale Mean if Item Deleted ATTRACT 236216 PARTRES 256486 PEOPLE 241892 MYAPP 238378 OTHERS 232162 PRIDE 238919 ILOOK 238919 ITRY 248649 Reliabili icients Alpha 1404 0822 1042 1317 1305 0607 3019 3064 mmwmmemm Correcte Item Total Correlation alpha Squared Multiple Correlation 2356 3077 0615 1357 1104 3076 2522 2238 1460 Alpha if Item Deleted 1090 0575 1292 0947 0558 0629 2115 5555 Item total correlations and alpha if deleted are all low Method 2 covariance matrix will be used for this analysis Example of Reliability Suggesting A Few Problematic Items R E L I A B I L I T Y A N A L Y S I S S C A L E A L P H A Mean Std Dev Cases 1 MEAPPR 35000 8602 260 2 MOREHR 20769 11974 260 3 CAREOT 33846 9829 260 4 FIRST 32308 11422 260 5 OWL 23462 10561 260 6 HOTDATE 31538 10466 260 7 NATURE 18077 5670 260 8 OTHERTTM 25385 11038 260 9 TREND 25385 13033 260 10 OWNSTYLE 24231 9454 260 11 SEXCON 33846 12985 260 Correlation Matrix MEAPPR MOREHR CAREOT FIRST KNOWL MEAPPR 10000 MOREHR 5048 10000 CAREOT 6623 4157 10000 FIRST 0814 0157 0110 10000 KNOWL 1101 1484 1334 2296 10000 HOTDATE 4443 1178 4068 2033 1308 NATURE 1230 1405 2208 2565 1824 OTHERTTM 5055 5727 4282 0244 1425 TREND 5709 3569 5188 3163 1207 OWNSTYLE 1230 2528 1622 1282 1926 SEXCON 3939 0317 1615 3693 1615 HOTDATE NATURE OTHERTIM TREND OWNSTYLE HOTDATE 10000 NATURE 0156 10000 OTHERTTM 1678 1721 10000 TREND 3767 1999 2631 10000 OWNSTYLE 5131 1406 1946 1648 10000 SEXCON 3962 0042 0451 6527 1053 SEXCON SEXCON 10000 R E L I A B I L I T Y A N A L Y S I S S C A L E A L P H A N of Cases 260 N of Statistics for Mean Variance Std Dev Variables Scale 303846 371262 60931 11 Item total Statistics Scale Scale Corrected Mean Variance Item Squared Alpha if Item if Item Total Multiple if Item Deleted Deleted Correlation Correlation Deleted MEAPPR 268846 298662 6934 6651 6746 MOREHR 283077 302215 4155 5095 7052 CAREOT 270000 308000 4913 6739 6960 FIRST 271538 318954 3043 3896 7220 OW 280385 346785 1071 3044 7472 HOTDATE 272308 318646 3526 6657 7143 NATURE 285769 360938 1043 4937 7370 OTHERTIM 278462 302154 4691 5268 6972 ND 278462 257354 7330 7696 6459 OWNSTYLE 279615 358785 0312 6580 7525 SEXCON 270000 293600 4321 6541 7029 Reliability Coefficients 11 items Weak items Alpha Standardized item alpha 7163 Method 2 covariance matrix will be used for this analysis Reliability After Three Weak Items are Dropped R E L I A B I L I T Y A N A L Y S I S S C A L E A L P H A Mean Std Dev Cases 1 MEAPPR 35000 8602 260 2 MOREHR 20769 11974 260 3 CAREOT 33846 9829 260 4 IRST 32308 11422 260 5 HOTDATE 31538 10466 260 6 OTHERTIM 25385 11038 260 7 TREND 25385 13033 260 8 SEXCON 33846 12985 260 Correlation Matrix MEAPPR MOREHR CAREOT FIRST HOTDATE MEAPPR 10000 MOREHR 5048 10000 CAREOT 6623 4157 10000 FIRST 0814 0157 0110 10000 HOTDATE 4443 1178 4068 2033 10000 OTHERTIM 5055 5727 4282 0244 1678 TREND 5709 3569 5188 3163 3767 SEXCON 3939 0317 1615 3693 3962 OTHERTIM TREND SEXCON OTHERTIM 10000 TREND 2631 10000 SEXCON 0451 6527 10000 N of Cases 260 0 Statistics for Mean Variance Std Dev Variables ScaIe 238077 320815 56641 8 R E L I A B I L I T Y Item total Statistics MEAPPR MOREHR CAREOT FIRST HOTDATE OTHERTIM TREND SEXCON Reliability Coefficients Alpha A N A L Y S I S S C A L E A L P H A Scale Scale Corrected Mean Variance Item Squared Alpha if Item if Item Total Multiple if Item Deleted Deleted Correlation Correlation Deleted 203077 251015 7239 6278 7296 217308 255646 4198 4466 7692 204231 255338 5619 5673 7474 205769 280138 2285 1879 7986 206538 260754 4594 3116 7619 212692 261246 4199 4077 7681 212692 217246 7127 6483 7132 204231 244938 4592 5787 7639 8 items Standardized item alpha 7910 NOTE First is how the weakest it but it does not need to be dropped as the scale as a whole is close enough This is an example of open by index that works The mean for index gets steadily higher for each category of open Report INDEX OPEN Mean N Std Deviation 100 230000 8 705084 200 329000 10 864356 400 344286 7 713809 500 436250 8 261520 Total 334242 33 981956 Report NEWINDEX OPEN Mean N Std Deviation 100 180000 2 00000 200 212857 7 160357 300 235000 4 57735 400 235000 8 346410 600 260000 1 Total 224091 22 288937 Report INDEX OPEN Mean N Std Deviation 100 262500 8 406202 200 310000 9 915150 300 362000 5 798123 400 433333 3 378594 500 422222 9 509357 Total 347059 34 907053 Test of Bivariate Association Between Index and Student Report NEWINDEX STUDENT Mean N Std Deviation 00 232353 17 253795 100 196000 5 230217 Total 224091 22 288937 ANOVA Table NEWINDEX STUDENT Sum of df Mean F Sig Squares Square Between Groups 51059 1 51059 8218 010 WithinGroups 124259 20 6213 Total 175318 21 ANOVA test is signi cant signi cance p lt 05 is conventional signi cance level The average on newindex for sex 1 is 196 versus 232 for sex 0 and the signi cance test says this is a large enough difference given the sample size that it is unlikely to be due to random chance TURN THE CONIPUTER PRINTOUT INTO A FINAL TABLE STUDENT Mean N Not a Student 0 232 17 Student 1 196 5 Total 224 22 POl Conclusion Students have a significantly lower score on Whatever the dependent variable is than nonstudents do FREQUENCIES VARIABLESFemale Year Wise Major Income White STATISTICSMEAN MEDIAN MODE ORDER ANALYSIS Frequencies DataSetl UDocumentsCoursesMethodstestdatasav Statistics Year in From Parents39 Female School Wisconsin Major Income White N VaHd 30 30 30 30 30 30 Mbgw 0 0 0 0 0 0 Mean 50 277 67 220 9666667 73 Median 50 300 100 200 9000000 100 Mode 0a 4 1 2 90000 1 a Multiple modes exist The smallest value is shown Frequency Table Female Cummmwe Frequency Percent Valid Percent Percent de Mab 15 500 500 500 Femab 15 500 500 1000 Tmal 30 1000 1000 Year in School Cummmwe Frequency Percent Valid Percent Percent de FwSh 5 167 167 167 Soph 7 233 233 400 Jr 8 267 267 667 Sr 10 333 333 1000 Tomi 30 1000 1000 From Wisconsin Cummmwe Frequency Percent Valid Percent Percent Valid Other 10 333 333 333 WiSC 20 667 667 1000 Total 30 1000 1000 Major Cumulative Frequency Percent Valid Percent Percent Valid Undecided 5 167 167 167 Social Science 5 167 167 333 Humanities 10 333 333 667 Natural Science 4 133 133 800 Business 3 100 100 900 Engineering 1 33 33 933 Agriculture 2 67 67 1000 Total 30 1000 1000 Parents39 Income Cumulative Frequency Percent Valid Percent Percent Valid 40000 1 33 33 33 50000 4 133 133 167 60000 2 67 67 233 70000 2 67 67 300 80000 4 133 133 433 90000 6 200 200 633 100000 4 133 133 767 120000 1 33 33 800 130000 1 33 33 833 150000 1 33 33 867 160000 1 33 33 900 180000 1 33 33 933 200000 2 67 67 1000 Total 30 1000 1000 White Cumulative Frequency Percent Valid Percent Percent Valid Minority 8 267 267 267 White 22 733 733 1000 Total 30 1000 1000 CORRELATIONS VARIABLESFemale Year Wisc Major Income White PRINTTWOTAIL NOSIG MISSINGPAIRWISE Correlations DataSetl Correlations UDocumentsCoursesMethodstestdatasav Year in From Parents39 Female School Wisconsin Major Income White Female Pearson Correlation 1 215 141 000 235 000 Sig 2tailed 254 456 1000 211 1000 N 30 30 30 30 30 30 Year in School Pearson Correlation 215 1 043 120 120 148 Sig 2tailed 254 820 529 527 435 N 30 30 30 30 30 30 From Wisconsin Pearson Correlation 141 043 1 000 454 107 Sig 2tailed 456 820 1000 012 575 N 30 30 30 30 30 30 Major Pearson Correlation 000 120 000 1 048 119 Sig 2tailed 1000 529 1000 802 530 N 30 30 30 30 30 30 Parents39 Income Pearson Correlation 235 120 454 048 1 360 Sig 2tailed 211 527 012 802 050 N 30 30 30 30 30 30 White Pearson Correlation 000 148 107 119 360 1 Sig 2tailed 1000 435 575 530 050 N 30 30 30 30 30 30 quot Correlation is significant at the 005 level 2tailed FREQUENCIES VARIABLESLegal Ondemand eright STATISTICSMEAN MEDIAN MODE ORDER ANALYSIS Frequencies DataSetl UDocumentsCoursesMethodstestdata Statistics Women39s Keep Abortion Abortion on Right to Legal Demand Choose N Valid 30 30 30 Missing 0 0 0 Mean 67 347 583 Median 100 400 600 Mode 1 4 8 Frequency Table sav Keep Abortion Legal Cumulative Frequency Percent Valid Percent Percent Valid N0 10 333 333 333 Yes 20 667 667 1000 Total 30 1000 1000 Abortion on Demand Cumulative Frequency Percent Valid Percent Percent Valid Strongly Disagree 4 133 133 133 Disagree 6 200 200 333 Neither Agree Nor Disagree 1 33 33 367 Agree 10 333 333 700 Strongly Agree 9 300 300 1000 Total 30 1000 1000 Women39s Right to Choose Cumulative Frequency Percent Valid Percent Percent Valid Extremely Unimportant 1 33 33 33 2 3 100 100 133 Unimportant 2 67 67 200 4 4 133 133 333 Mildly Unimportant 1 33 33 367 Mildly Important 5 167 167 533 7 5 167 167 700 Important 6 200 200 900 9 2 67 67 967 Extremely Important 1 33 33 1000 Total 30 1000 1000 GRAPH HISTOGRAMeright Graph DataSetl UDocumentsCoursesMethodstestdatasav 5 L I Frequency 1 I I I I I I 0 2 4 6 8 10 12 Women39s Right to Choose CROSSTABS TABLESFemale BY Legal FORMAT AVALUE TABLES STATISTICCHISQ CELLS COUNT ROW COUNT ROUND CELL Crosstabs DataSetl U DocumentsCoursesMethodstestdata sav Case Processing Summary I Cases I I Valid Missing Total I N I Percent I N I Percent I N I Percent I Female Keep Abortion Legal 30 I 1000 0 I 0 30 I 1000 Female Keep Abortion Legal Crosstabulation Keep Abo tion Legal No Yes Total Female Male Count 7 8 15 Within Female 467 533 1000 Female Count 3 12 15 Within Femaie 200 800 1000 Total Count 10 20 30 Within Femaie 333 667 1000 ChiSquare Tests Asymp Sig Exact Sig Exact Sig Value df 2sided 2sided 1 sided Pearson ChiSquare 2400b 1 121 Continuity Correctiona 139350 1 39245 Likelihood Ratio 2451 1 117 Fisher39s Exact Test 245 123 LinearbyLinear Associaton 2320 1 128 N of Valid Cases 30 a Computed only for a x2 table b 0 cells 0 have expected count less than 5 The minimum expected count is 500 MEANS TABLESOndemand BY Year CELLS MEAN COUNT STDDEV STATISTICS ANOVA Means DataSetl U DocumentsCoursesMethodstestdata sav Case Processing Summary on Year in School 30 1000 30 1000 Report Abortion on Demand Year in School Mean N Std Deviation Fresh 400 5 1225 Soph 357 7 1512 Jr 363 8 1768 SF 300 10 1333 Total 347 30 1456 ANOVA Table Sum of Squares df Mean Square F Sig Abortion on Demand Between Groups Combined 3877 3 1292 584 631 Year In School Within Groups 57589 26 2215 Total 61467 29 Measures of Association Eta I Eta Squared I Abortion on Demand Year in School 251 063l REGRESSION MISSING LISTWISE STATISTICS COEFF OUTS R ANOVA CRITERIAPIN05 POUTlO NOORIGIN DEPENDENT eright METHODENTER Income Regression DataSetl U DocumentsCoursesMethodstestdata sav Variables EnteredRemovedb Variables Variables Model Entered Removed Method 1 Parents Enter lncomea a All requested variables entered b Dependent Variable Women39s Right to Choose Model Summary Adjusted R Std Error of Model Square the Estimate R R Square 1 288a 083 050 2360 a Predictors Constant Parents39 Income ANOVAb Sum of Model Squares df Mean Square F Sig 1 Regression 14162 1 14162 2542 122a Residual 156004 28 5572 Total 170167 29 a Predictors Constant Parents39 Income b Dependent Variable Women39s Right to Choose Coefficientsa Unstandardized Standardized npff39 ipnt npffir ipnt Model B Std Error Beta t Sig 1 Constant 7395 1070 6911 000 Parents39 Income 162E005 000 288 4594 122 a Dependent Variable Women39s Right to Choose Sociology 357 Methods of Sociological Inquiry Lecture Notes 1 Basic Methodological Concepts Getting Started I Why bother to learn hoW to do research 7 Answer questions important to us 7 Informed consumer of others research I Empirical statemenw 7 Possible to be true or false 7 Truthfalsehood can be determined by observation I Values 7 what ought to be what is good or bad 7 Distinguished from empirical statements 7 In uence se1ection ofresearch topics Nonscienti c Approaches to Knowledge I Unsupported assertions 7 no evidence at all I Appeals to authority 7 Need to evaluate the authority 7 Have to do this sometimes I Casual observation 7 Unconcretized abstractions eg aggressive 7 speci c behaviors unclear 7 Errors of observation Errors of Observation I Inaccurate observation Simply wrong about what you think you saw I Overgeneralization Correct about what you saw but apply it too broadly I Selective observation Only notice what supports your view Scientific Approach to Knowledge I Empiricisrn take sensory input from the world I Objectivity different people can observe the same inputs and see the same 39ng 7 Reliability observations can be consistently made over time and by different people about observation 7 Validity theoreticalconceptual interpretation of observations is correct about link between observation an theory I Controlled observation to eliminate sources of error in observation I Knowledge as tentative subject to further re nement Scienti c Attitude I Knowing better than not knowing About things oumide you not introspection I Respect the evidence prefer unpleasant truth to selfdelusion I Always tell the truth about your research no matter what I Recognizes the boundaries of expertise Studying People I Free Will Vs social patterns 7 Social constraints on choices e g rnoney laws 7 Social effects on choices lead people to want to behave in certain ways I Probabilistic thinking 7 Recognize both the general tendency and the variation around that tendenc I Understanding Verstehen Vs prediction exp anation The Scienti c Process predictions hypotheses Empirical Generalizations l Theory I Not idle speculation but coherent accounts of how things Work I Theory is part ofscience making sense ofwhat the distinct observations mean how they are related I Abstractions are concretized theoretical elements have connections to empirical phenomena I Explanation and prediction as equivalent Goals of Scienti c Research I Description Accurate description of What is actually happening is the bedrock of science 7 Comparative description documents changes over time or di erences between subjects Explanation and Prediction Provide a theoretical account for observations as part of a larger understanding of how the World Works and predict future observations Understanding Of people Why they do What they do Ofsocial systems how they Work and Why they Work that Way Research Steps De ne a problem of interest Identify empirical questions Within that problem Usecreate a general theoretical framework for understanding the problem Concretize the concepts in the theory 7 operationalize the variables De ne the sample Collect data Analyze data Draw conclusions More examples of units of analysis and variables Unit of Analysis Variables Objects Characteristics of objects which vary Individuals income age sex attitude toward abortion how voted in 2000 Households income not the same as individual income size number ofpeople marital status of head Organizations size number of organizational levels sex composition Census tracts median income percent Hispanic proportion of dwellings owneroccupied Metropolitan areas population percent Hispanic number of jobs in manufacturing region of the country in which the metro area is proportion of population who are college graduates Nations population GNP percent of population with access to electricity system of government
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'