Comm88 - All Lecture Notes Spring 2014
Comm88 - All Lecture Notes Spring 2014 Comm88
Popular in Communication Research Methods
Popular in Communication Studies
This 61 page Bundle was uploaded by Kelsey Calef on Thursday October 2, 2014. The Bundle belongs to Comm88 at University of California Santa Barbara taught by Dolly Mullin in Spring2014. Since its upload, it has received 111 views. For similar materials see Communication Research Methods in Communication Studies at University of California Santa Barbara.
Reviews for Comm88 - All Lecture Notes Spring 2014
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: 10/02/14
Communication 88 Comm Research Methods Lecture 4114 Objectives lreparation for the major lreparation for consuming and creating knowledge Research paper ltart early with group do things in order do NOT skip literature review llnderstand project Practice thinking speaking like scientist Lecture 4314 Ways of Knowing Aside from research how do you know some truths 0 IE Vegetables are good for you I Possible sources parents doctors studies media experience gov39t agencies ideologies 0 IE People who are similar to each other tend to like each other I For example values about money are very important thrifty vs expensive jays of knowing epistemology science of knowing 0 What does it mean to know something 0 Everyday ways of knowing and their problems I Method of tradition tenacity Get carried down passed on over and over even if origin is unknown Truths are persistent pieces of knowledge don39t just go away Upside Do not have to do research when acquiring knowledge this way 0 IE Supreme court using precedent to decide what cases to judge Downside Might not be true or best interpretation 0 IE Waiting 30 minutes after eating to swim I Method of authority Taken to be true because some granted authority said so 0 IE surgeon general teacher parent professor doctor Upside Don39t have to research saves time has experience and professionalism Downside Not knowing if source is reliable can be wrong I Method of experience observation Know it39s true because seen with own eyes 0 Personal experience easy superficial noticing I IE Remember when I got food poisoning at that Taco Bell 0 Baconian Empiricism serious rigorous empirical experiments I IE Biohacking when chemists use their own bodies to test intake effects I IE If milk is past expiration label authority smelling it to check if it39s good Downside Could be situational or inaccurate observation misunderstood or selective observation noticing one pattem ignoring others I Method of intuition logic Gut feeling able to reason to get to truth common sense Platonic idealism Plato argued that you can39t trust your eyes but your brain by setting up series of logical premises to get to truth If A is B and B is C A is also C Downsides One39s common sense could differ from another 0 IE Both sides of politics believe theirs is common sense approach 0 What if A doesn39t actually equal B If premises are incorrect from start argument will still be wrong 0 Illogical reasoning All sh can swim I can swim so I am a sh Too many Bs on a test must be wrong 0 Problem for all of the above methods Overgeneralization I Kemel of information taken to mean pattem I IE Food poisoning at Taco Bell basis for never eating at any Taco Bell or Mexican again 0 Everyday ways of knowing can even lead to con icting ideas about trut I IE Absence makes the heart grow fonder or Out of sight out of mind or reconciling Absence extinguishes small in ames the great All can be con rmed with methods but are contradictory The Scienti c Method 0 Attempt to control or overcome problems in attaining knowledge 0 Combines platonic idealism with Baconian empiricism I Logic intuition gt constructing theories I Observation experience gt gathering data In communication science Use empirical observations to test theories about comm pI39OC SS S Unique characteristics of science 0 How is science different from other everyday ways of knowing I Scienti c research is public Published in peerreviewed journals O Submit study to journal have it checked by others in area of study make advised revisions can be reviewed again then possibly published lpportunity to replicate studies 0 IE Chemists went to media claiming ability to create cold fusion but other researchers could not replicate study Section 4414 How to read the accompanying text 0 Use to sum up and in other words in short etc and describing cues notes importance 0 Quizzes and outlines of chapters available online I Don39t get bogged down with examples skim for key points I Portions like field research not as important for quantitative class as opposed to empiricism reasoning etc Lecture 4814 Unique characteristics of science continued 0 Science is empirical llore than simply noticing My friends who eat vegetables are healthier I Rigorous empiricism conscious deliberate observations let up study with controlled factors 1lany observations are made 0 IE more studies repeated results can see a pattem 0 Science is objective I Opposed to subjective which is biased Control remove personal biases Even so there is always a spin to objective research 0 Public side attempts to catch those viewpoints must be able to challenge Explicit rules standards amp procedures 0 Science is systematic amp cumulative I Builds on prior studies theory I Practically impossible to do totally original research Must ask what we know already I New knowledge modifies old 0 Goals of scientific research What can science tell us I Description Looking for social regularities of aggregates 0 Means seeking pattems things that happen regularly 0 Science not interested in a unique case I IE predictors of infidelity to con rm evidence of pattems I Aggregate deciding what factors will be used in research IE infidelity in freshman females I Science can tell is what is Handles description fairly well I Explanation Develop understanding of WHY pattems exist what causes what Can create a causal chain 0 Science can tell us why it is I Prediction Predict outcomes given certain factors If I know X Y Z I can tell what you39re likely to do Science can tell what will be within certain realm of probability 0 Can sometimes make good prediction without an explanation I Science CANNOT settle questions of value CANNOT tell us what should be right wrong good bad moral immoral Science get brought into debates but cannot address rightness or wrongness 0 IE abortion debate science can tell fetus development is used to support or refute both sides of debate but not if abortion should be considered okay 0 Science can only inform debate The Research Process Theories Hypotheses amp Research Questions 0 Wheel of Science cyclical process of research always ongoing building of knowledge 0 Theories gt Hypotheses gt Observations I In this order engaging in deduction Traditional science quantitative methods I Theories Something scientists posit may be true having reason to suspect 0 Can also be Observations gt Empirical generalizations gt Theories I In this order engaging in induction Humanistic interpretive qualitative methods To remember difference think of how gathering data inward to make point inductive Difference between quantitative and qualitative 0 Quantitative I Employ numerical measures amp data analysis I Adhere strongly to scientific goals amp principles objectivity empirical data etc I Examples Surveys Experiments Content Analysis 0 Qualitative I Also called interpretive research or field research I A humanistic form of social science Lalues some aspects of science especially empiricism 0 Data is very significant Having personal spin on it is okay 1alues researcher subjectivity I Examples Participant Observation Depth Interviewing Conversation Analysis I Note There is also purely humanistic research in communication called critical studies IE rhetorical criticism feminist analysis cultural studies etc What is the difference between Basic and Applied Research 0 Basic theoretical Research I Testing building theories conceptual ideas advancing what we know about a topic Laluable aside from having practical use LE Math is valuable for its own sake 0 Applied practical Research I Applying research to solve practical problems I IE testing effects of an ad campaign policy change new school program company technology etc Use math to build bridge 0 Matter of emphasis in any given study these descriptions are of extremes I Note Even the most theoretical research has practical value and even most applied research uses theoretical reasoning and arguments to form hypotheses etc Lecture 41014 Using Theories in Research 0 Theory an attempt to explain some aspect of social life I A scholar39s ideas about howwhy eventsattitudes occur I Includes set of concepts and their relationships Terms for thingsideasparts of the theory because they are conceptualized Researchers must define these terms 0 IE Social Cognitive Theory Bandura I Theory about how we leam mostly applied to children or in communication media effects I Theory that leaming happens through watching modeled behavior see parent brushing teeth child imitates Requires attention retention motor production motivation e g rewardspunishments 0 What are some concepts involved here I What does it mean to retain information I What is it to be modeled I Does it have to occur inperson I Definition of rewardpunishment I Definition of leaming etc I Concepts are studied as variables They have variations that can be measured 0 IE Motivation I How does it vary Rewarded vs punished model amount of rewardpunishment etc 0 IE Model I TV character vs parent TV hero vs villain degree of likeability or similarity to viewer etc 0 Scienti c theories should be falsifiable I If not a study cannot actually test the theory I Able to be tested empirically with data I There is some result that if you got it would show the theory is wrong I You are taking a stance on one side of a theory Note you can never prove theories true can only gain supportevidence 0 IE What about Soc Cog theory I Has been tested many times cannot say it39s right only that there is much evidence for it 0 What about theory of man made global warming I Evidence suggested in press temp change ice change droughts oods lackexcess of hurricanes etc many contradictions I Ends up being politics and less of a scientific theory Using Theories in Research 0 From prior ndings andor theory we derive a testable hypothesis I A speci c prediction about the relationship between variables in your study I IE Using Soc Cog Theory to make a prediction about the effects of TV violence H1 TV violence viewing will produce more aggressive behavior than will non violent TV viewing What are the variables involved here 0 Definition of violence definition of aggressive behavior etc I What if theory or previous research does not lead to a specific prediction Or if previous findings con ictinconclusive Pose research question instead of hypothesis 0 Difference is that research question is not a statement but an actual question 0 IE To what extent will children imitate behavior of a TV character whom they do not likerelate to Will there be gender differences in children39s imitation of violence I If study has only research questions appears to be a lack of background knowledge Testing a hypothesis An example 0 Researcher A I Soc Cog Theory children leam behavior by watching models behave I Hypothesis watching TV violence will increase kids aggressive behavior I Given large grant to conduct research Finds out how much violent TV viewed How much aggression is shown on playground Random sample of subjects used Each scored and plotted on X Y axis shows positive linear relationship 0 Conclusion TV violence increases aggression I Possible problems Correlation relationship bw variables is not causation Could be another cause for aggression in subjects perhaps parental neglect leading to more violent viewing leading to aggression Cannot tell difference from aggression before and after chickenegg issue Problem with conclusion TV violence increases aggression gt should be TV violence is related to aggression 0 Researcher B I Catharsis Theory watching others behave allows purging of pentup feelings I Hypothesis watching TV will reduce kids aggressive behavior I Lack of funds for research llas subjects watch one of four clips 0 5 10 20 acts of violence in clip of varying degrees of violence jandom assignment of subjects 11 umber of hits on toys variety available without other kids toys used as weapons jot on XY axis acts of aggression with degree of violent content viewed negative linear relationship 0 Conclusion TV violence decreases aggression I Many possible problems with lack of resources I However conclusion correct because controlled study accounts for behavior prior to viewing Technically should state correct for these lab conditions Often tend to give up generalizability for control vice versa for A Types of Hypotheses amp Research Questions 0 Hyps and RQs can be I Causal state how 1 variable changesin uences another Effects of something on something else effects of violent tv produces violent behavior I Correlational state mere association between variables lhings go to variables the more tv violence children watch the aggressive they will be OR aggressive kids watch more violence than nonaggressive kids I See pg 107 on continuous interval ratio and difference categoricalnominal statements Section 41114 Recap llniqueness of Science Objectivity of process questions aren39t necessarily objective but process is limpirical observations research lystematiccumulative in organization iAvailable publicly Science does lixplain jescribe Predict Science cannot Judge make evaluations InductionDeduction Induction qualitative gathering data inward bottomup theories come last Deduction quantitative topdown most common for communication methods theories first Lecture 41514 Variables 0 Independent xed predictor I In case of Exercisel example Hypothesisl is independent variable 0 Dependent Outcome I In example the scores of student recall Different methods for different hypotheses 0 SurveyCorrelational Research eg Researcher A I Test correlational hypothesesRQs mere relationshipassociation Measure some variables amp relate them compare existing groups etc I Great for extemal validity Ability to generalize results to other people andor normal life settings I Poor for causality Measuring factors simultaneously cannot tell which came first 0 Experimental Research eg Researcher B I Tests causal hypothesesRQs Manipulate vars grounds control everything else and measure effects eg Researcher B manipulated amount of violence seen I Great for intemal validity basically means causality Ability to establish that X causes Y rules out other explanation I Poor for generalizability The Research Process cont De ning concepts and variables Variables in Experimental Research causal hyps 0 Independent Variable IV I The cause in causeeffect relationship I Variable manipulated by researcher 0 Dependent Variable DV I The effect or outcome I Variable affectedchanged by the IV lixample hypothesis 0 Greater physical attractiveness will create impressions of greater friendliness I Is suggesting causal relationship LV physical attractiveness eg manipulate level of attractiveness LDV impressions of friendliness eg ratings on friendliness scale I Can39t be causeeffect so LV considered a predictor variable V is what is being predicted by the IV sometimes called criterion variable lixample hypothesis 0 Stronger fan identity predicts greater participation in online fan forums I IV fan identi cation eg score on identi cation scale I DV fan forum participation eg measure how often people post or report reading posts etc I Could the IvsDVs be other way around Yes Not manipulating relegating cause but as researcher get to decide which is which depends on how study is set up De ning ConceptsNariables 0 Conceptual de nition I A working de nition of what the concept means for purposes of investigation usually based on theoryprior research Example variable fear what is it 0 Feeling unsafe reaction to threat possibly evolutionary in nature adrenaline rush from euphoria of not dying needs to be de ned further 0 Operational de nition I How exactly the concept will be measured in a study Eg fear de ned as sharp heart rate increase combined with nonverbal reaction Types of Measures 0 Physiological measures I IE BP brain imaging Cortisol stress hormone even glycosulated hemoglobin scores for diabetes 0 Behavioral measures I IE Nonverbal gestures timemoney spent actual posts on social media 0 Selfreport measures I IE items on questionnaire Lecture 41714 Levels of Measurement Nominal categoricaldiscrete 0 Variable measured merely with different categories works best with separate groupings I IE yes or no Qs gender mf ethnicity TV violence rewardedpunished TV use hilo 0 Categories must be mutually exclusive Subjects cannot pick both I check all that apply just a way of asking many YN questions 0 Categories must be exhaustive All possible options must be offered should not force agreement with wrong answer that doesn39t re ect subject39s information 0 Nominal measures are for comparing differences Ordinal 0 Variable is measured with rank ordered categories I IE Rank top 5 favorite TV shows rank most to least important political issues I not very useful compared to Nominal Interval 0 Variable is measured with successive points on a scale with equal intervals aka equidistant points I IE Olympic medalists may win Silver but could have been only 1 point behind gold no way to tell how far apart Interval accounts for differences on continuum IE measure on immigration policy opinion US should build a fence along border Strongly oppose l 2 3 4 5 strongly favor llumbers themselves are just relative to each other Strongly oppose 2 l 0 1 2 strongly favor same scale as prior above 0 0 does not mean nothing here still represents a measurement vs ratio Ratio 0 Interval measurement with a true absolute zero point I Nothing has been measured yet absence of data I IE Time in hours weight in lbs age in years etc I IE Test scores if from 0 possible SAT not ratio bc you get points for signing name 0 means nothing in all these cases Interval and ratio measures are continuous variables vs difference of nominal and ordinal 0 Allow continuoustype hypotheses the more X the more Y etc I IE misunderstanding in This Is Spinal Tap with amp going to 11 Measures should capture variation 0 Use continuous vars for DV39s where possible have good conceptual fit with variables in the HypsRQ39s 0 Conceptual and operational definitions should be answered or reasonable to attempt by measurement minimize potential social desirability effects 0 Lying exaggeration political correctness Using Questionnaire Items as Measures Common for IV39s and DV39s in surveys Common for DV39s in experiments IV is a manipulation Types of Questionnaire Items 0 Openended I Respondents give their own answers to Qs 0 Closedended I Respondents select from list of choice exhaustive and mutually exclusive Some closedended formats 0 Likerttype items I item just 1 unique thing on survey or to make up parts of variable not always Q I Respondent indicate their agreement with a particular statement LE Parents should talk openly about sex with their children Strongly disagree 1 2 3 4 5 Strongly agree liave to know what you want high scores to represent can format results in such a way lther response options also possible opposefavor not at allvery much almost neveralmost always 0 Semantic Differential I Respondents make ratings between two opposite bipolar adjectives I IE My best friend is IntelligentUnintelligent I Can reverse code numbers to make easier to read again what you would want your high scores to re ect IE high score warmer more intelligent Section 41814 include limitations of studies in data jwo different kinds of Validity 0 External generalizability in surveys 0 lntemal in experiments Career day This Sat Corwinn pavilion at noon QA session Monday 28 Lecture 42215 Composite Measures Grouping conglomerate form of other measures Use multiple items for one variable combine those items into an index aka scale 0 Example variable Perceived credibility of a speaker I IE Professional dress vs hipster style 0 As a singleitem measure I The speaker I just heard is credible 765432l not credible I Above semantic differential measurement I Likert The speaker I just heard is credible Agree or disagree Semantic differential works better bc don39t need statement each time 0 As a composite measure I The speaker I just heard is credible 7l not credible knowledgable 7l not knowledgable experienced 7 l inexperienced honest 7l dishonest etc I Here all high scores mean more of each item 0 Unidimensional index all items added or averaged into E overall score I Options Unidimensional credibility add all items into one total credibility score Multidimensional credibility knowledge experience competence Expertise dimension trustworthiness honesty unbiased Trustworthiness dimension 0 Multidimensional index group different items into different subscales I Separating all the different dimensions of a Variable How Good is Your Measurement Reliability and Validity Reliability of Measurement 0 Are you measuring the concept consistently 0 For measures using questionnaire items I Interitem reliability Administer same items more than once e g testretest splithalf 0 Can get another set of subjects from same population get same subjects to take survey twice Look at intemal consistency of similar items in a scaleindex e g Cronbach39s alpha 0 Similar items should get similar scores 0 Cronbach39s alpha Numerical formula for finding whether interitem reliability is good or not good usually above 7 I Unidimensional credibility is likely to get LOW Cronbach39s alpha poor quality I Multidimensional credibility is likely to get higher reliability bc computed separately for each subscale Expertise dimension and Trustworthiness dimension 0 For measures using coders eg behavioral observations I Intercoder reliability Compare multiple coders I lntracoder reliability Compare multiple observations of same coder Validity of Measurement 0 Does your measure really capture the concept you intend to be measuring 0 Assessing validity I Subjective types of validation 3 ace validity 0 The measure lookssounds good on the face of it Content validity 0 The measure captures the full range of meaningsdimensions of the concept Criterionrelated validation aka Predictive Validity 0 The measure is shown to predict scores on an appropriate criterionfuture measure 0 Example SAT scores your potential to achieve gt college GPA your achievement Construct validation 0 The measure is shown to be related to measures of other concepts that should be related and not to ones that shouldn39t 0 EX Verbal aggressiveness scale lt gt hostility scale FYI don39t worry about convergent discriminant Can a measure be reliable but not valid Yes Can a measure be valid but not reliable No If not consistent how can it be true Lecture 42414 Sampling How we select participants or other units for a study Sample A subset of the target population whowhat you want to report about IE teenagers college students voters FB users married couples juries football fans etc IE content analysis of TV shows magazine ads blog posts etc jepresentative sampling probability sampling 0 Intended to be miniature version of the target population 0 KEY is random selection I Everyone in population has equal chance of being included in sample I How representative is it 0 Will always be Sampling Error I Error variation Sample data will be slightly different from population because of chance alone AKA random error ltatistically known as the margin of error 0 IE National poll N 1000 gt 3 larger sample size smaller margin of error Representative Sampling Techniques 0 Simple random sampling I Select elements randomly from population Listed populations random 39s table lhones randomdigit dialing 0 Systematic random sampling I From a list of the population select every Nth element AND I Must have random start cycle through entire list Similar results as SRS 0 Example School population 60 girls 40 boys Pull 8 out of 24 names random start then every 3 Comes close to 60 40 Watch out for potential periodicity 0 Strati ed sampling I For getting population proportions even more accurate I Divide population into subsets strata of a particular variable Usually strati ed for demographic variables e g sex race political party I Select randomly from each strata to get right proportions of the population I Need prior knowledge of population proportions I Increases representativeness bc reduces sampling error for the strati ed variables I But more costly amp time consuming 0 Multistage cluster sampling I Useful for populations not listed as individuals First randomly sample groups clusters then randomly sample individual elements within each cluster lixamplez Sampling high school athletes 15 stage Random sample high schools 2 stage Random sample athletes from those schools in sample Reduces costs But sampling error at each stage So for all of Representative Sampling techniques 0 Will always have sampling error 0 But can generalize to the larger target population assuming done properly 0 Caution Avoid Systematic Error I Systematically over or underrepresent certain segments of population I Caused by improper weighting Very low response rate jrong sampling frame Using nonrepresentative sampling methods Nonrepresentative sampling cannot generalize to a population 0 Convenience sample I Selecting individuals that are availablehandy 0 Purposive sample Select certain individuals for special reason their characteristics etc 0 Volunteer sample I People select themselves to be included 0 Quota sample I Select individuals to match demographic proportions population I Differs from stratified in that individuals are not randomly chosen 0 NetworkSnowball sample I Select individuals who contact other similar individuals and so on POST MIDTERM Lecture 5114 More Q39s for Final from this day BC low attendance Survey Research Primary Goals 0 lldentifydescribe attitudes or behaviors in a given population 0 lEXamine relationships between the attitudebehavior variables measured I Does X predictrelate to Y Ex Does exposure to alcohol ads X predict teen drinking Y Tricky bc frequent drinkers could know ads better peer pressure in uence etc I Do many factors together predict Y EX Do alcohol ads Xl parent drinking X2 peer drinking X3 and risk taking X4 together predict teen drinking Y 0 All factors related Administering Surveys 0 lSelfadministered questionnaire I Use of term survey refers to method actual questions are on questionnaire as items I Mail surveys online or e mail handouts diaries Relatively easy and inexpensive No interviewer in uence Increased privacyanonymity BUTl Must be selfexplanatory Example Media Use Diary is overly complicated hard to understand arduous 0 Can skew data Very low response rate of selfadmin Ways to increase response rate 0 Have inducements incentive 0 Make it easy to complete and retum 0 Include persuasive cover letter andor do advance mailing I Substantially increases likelihood of response 0 Send followup mailings llnterview Surveys 0 More exible can probe for depth 0 Higher response rate BUT 0 More potential for interviewer in uence 0 Higher costs Telephone 0 Quickest results 0 Compared to facetoface reduced costs more privacy more efficiency 0 Compared to selfadmin more detail possible better response rate 0 But what about call screening and cell phones Role of time in surveys lCrosssectional studies 0 One sample at one point in time IE each variable is measured once lLongitudinal studies 0 More than one point in time measured 0 Variables measured more than one time to track changes over time I Panel same people each time I Trend different random samples from same population ge g survey Americans every 5 years to study churchgoing poll lively voters over the course of an election campaign student alcohol study Samples do need to be representative I Cohort different samples but of same cohort l Here group of people are anchored to each other because of something in time timebased characteristic IE when you39re bom when you graduated lived in NY during 9 l 1 ge g survey class of 2012 every 5 years to check their employment since graduation Lecture 5614 How Do We Relate Variables in Survey Research Recall goals of survey Identifydescribe attitudes or behaviors in a given population Examine relationships between attbeh variables measured 0 Does X predictrelate to Y I Ex exposure to alcohol ads X predicts teen drinking Y 0 Do many factors together predict Y I EX Do alc ads Xl parent drinking X2 and peer drinking X3 predict teen drinking Y Depends on hypRQ and how your variables are measured When both IV amp DV are nominalcategorical discrete variables 0 All you can do is break down the percentages by category I EX YesN o MF supportopposeno opinion 0 Percent of people in 1 category who are also in another I EX Gallup survey on support for legalization of marijuana Q Do you think the use of marijuana should be made legal or not 44 Yes 54 No Could this be related to another variable How about gender Ex RQ Does gender IV predict support for legalization DV 0 By 2009 gender connection gone both closely equal in support 0 Political ideology makes difference I Does not address strength of opinion If IV is categorical but DV is intervalratio data continuous 0 lDifference statement in hypothesis I DV uses Likert semantic diff items etc 0 Compare mean average DV scores for the different IV categories I Ex Average of heavy teXter39s self esteem compared to average of light teXter39s self esteem I Same procedure as experimental where causality is added 0 Comparing means Ex RQ Does political ideology IV predict support for legalization DV LV political ideology 0 Measured as categoricalnominal variable I consider myself check one Liberal Moderate Conservative 0 OR if IV originally measured as a continuous var but then collapsed to categorical I Ex I consider my political views to be Very liberal l 2 3 4 5 6 7 Very conservative I Must decide how high is a high score Above median conservative below median liberal DV support for legalization 0 As continuous var intervalratio I The recreational use of marijuana should be made legal Strongly agree 1 2 3 4 5 6 7 Strongly agree To relate vars For Ss in each IV category compute their mean score on the DV then compare means 0 EX Conservatives M 23 Moderates M 42 Liberals M 61 I Liberals are significantly more in favor of legalization Signi cant statistical test shows 6139s difference from 42 did not occur by chance 0 All of stats just asks if something happens by chance If both IV and DV are intervalratio data 0 lCompute a correlation I Statistical value that relates two or more continuous variables I Compute r value Pearson r r tells you type vs and magnitude strength of relationship 0 Type of relationship I Positive r as X increases Y increases Line looks like Also called direct relationship I Negative r as X increases Y decreases Line looks like Also called inverse relationship 0 Magnitude of relationship I r ranges from 0 to 1 100 lt 0 gt 100 I The further from 0 the stronger the relationship Lecture 5814 Survey research cont What can you conclude from surveycorrelational data CAN conclude that Variables are relatedassociated CANNOT conclude that one Variable causes the other 0 Why not 0 Remember To establish causality I Variables must be related X correlated with Y Okay so far surveys can show that e g increased time studying GPA increases increase of thin mag exposure decrease body image Must establish time order 0 IV happened before DV Must rule out other explanationscauses 0 lSo SurveyCorrelational Research has 2 causality problems I lCausal Direction Problem time orderl Does X cause Y or does Y cause X Chicken and egg problem I Third Variable Problem Does some 3rd Variable explain the XN relationship Getting closer to causality 0 To help solve 3 Variable problem I l Partial correlation Measure potential 3 Variables Statistically partial out control for effects of those 3 Variables Then see if XN relationship still holds Partial correlation like Venn Diagram overlap of Variables 0 3 variable can also overlap 0 Can mathematically block out this part of overlap once controlled for I Check if anything is left over 0 If XN rel still holds can rule out 3 variable as the cause I If 3 has lots of overlap the correlation of XN relationship disappears or is reduce substantially then the 3 variable explanation matters I 3 could be what is driving 2 variables even if they are related 0 To help solve causal direction problem I Need a longitudinal studyl Helps avoid time order problems Crosslagged panel design 0 Time 1 Measure X and Y variables O Time 2 Measure X and Y vars again later for the same people 0 Compute r39s for X and Y but across the times measured I Example Age 8 Timel Age 18 Time2 TV violence X lXl Set of scores Aggression Y Y2 Scores r 31 r 0 no correlation So measure both variables for same people at different times then see which cross relationship holds 0 If Xl correlates with Y2 can get closer to supporting stronger intemal validity 0 Example of recent study Effect of IM on adolescents friendships IV IM use DV friendship quality I Signi cant r for IM Use at time 1 and Quality of Friendship at time 2 I No signi cant r for Quality of Friendship at time 1 and IM Use at time 2 Mediating Variable intimate selfdisclosure in IM Research claim errors example 0 IE BREAD ALERT A recent headline read Eating or smelling baked bread may be hazardous to healt I Some breadrelated research claims 1lore than 98 of convicted felons are bread eaters 1lore than 90 of Violent crimes committed with 24 hours of eating bread lully half of all children who grow up in breadconsuming households score below average on standardized tests Gateway food item leading to pbj cold cuts 0 Researchers make these same mistakes when proposing data results with less ridiculous data in emotionally charged cases especially I All of these prior claims have way too many 3 Variable problems Lecture 51314 lEXperimental Research Purpose To test hypotheses of cause and effect Goal is to establish intemal Validity 0 IV has effect on DV only through control Willing to sacri ce extemal Validity 0 Remember to establish causality intemal validity I Variables must be related I Must establish time order I Must rule out other explanationscause Key Elements to a True Experiment 0 Manipulation of causal variables I Divide Independent Variable IV into conditions Ex IV New painkiller drug half of Ss get drugother half do not while controlling all other variables Ss in each condition treated the same etc Examine effects on Dependent Variable Ex DV amount of perceived pain eg likert scale 0 lRandom Assignment of participants to conditions I Everyone must have equal chance of ending up in either condition Could be assigned to either condition to each control non group by chance I Why important Makes groups equal before manipulation O From book Matching can be used with RA not as substitute I Subjects are paired with those who score similarly on variables related to experiment are then randomly split into control or noncontrol group True Types of Experiments 0 Design notation X IV manipulationtreatment 0 Observation measure for DV I lPosttest only control group design R X 01 group 1 R Ol group 2 Example R R DV measure Compare them O1 beliefs about smoking O1 beliefs about smoking Occurs only after subjects are shown ad could realize goal of experiment If you get a statistically signi cant difference between group means on 01 the IV caused it Variations More groups several different treatments Example IV different types of ad appeals R Xl personal cancer story Ol group 1 R X2 cancer stats Ol group 2 R X3 tobacco industry 01 group 3 I lPretestposttest control group design R R 01 X 02 group 1 O1 02 group 2 Example R Ol smoking beliefs K antismoking ad 02 smk beliefs R or smoking beliefs 1 no antismoking adl o2 smk Beliefs I f T J L T lt39gti DV measure before Again if difference between group means on 02 the IV caused it Possible problem Differences on 02 could result of interaction of manipulation with pretest Looking for change with pretest I Soloman fourgroup design 2 groups both take pretest 1 gets treatment other doesn39t The other 2 groups do posttest only 1 w treatment 1 wo Can test interaction of pretest with treatment controls for change in groups R X Ol grpl Lecture 51514 Threats to Intemal Validity From book Ways that Variables other than IV could cause a change in DV If NOT a TRUE experiment or if do experiment improperly then gt 0 Alternative explanations become possible IE threaten internal Validity 0 Missing RA not even second group treats all in study same way only one period of time for one group examined no comparison 0 IE when a teacher wants to test if you smile the world smiles with you so has students smile all day and take notice of what happens I Could be many altemative explanations Smile at people already smiling could be lovely day there was free cake need control group by smiling at some not smiling at others lOne group pretestposttest design 01 X 02 group 1 0 Like above only includes a nosmiling group but now have a before and after two points in time no control group 0 Possible altemative explanations Could be seasonal like for ads and sales black Friday natural uctuation from time 1 to time 2 time of day happen to already be smiling lStatic group comparison posttest only nonequivalent groups X 0 1 group 1 01 group 2 0 Have 2 separate groups but neither is controlled no random assignment 0 Go to one half of campus to smile other half of campus to not smile compare groups 0 Possible altemative explanations People in each group could be different already Threats Within preexperiments Selection bias 0 Altemative explanation is because of people selected or how they are divided 0 All 3 preexperiments aka quasi experiments have this problem 0 True experiments use RA lHistory effect 0 Something that happens in world outside of control of data is producing results I Can be as little as a sunny day or free cake day does not have to be huge event could be happing personally to a participant 0 True experiments needuse RA to make sure it is happening to 1 group then the other I Equal amounts of people in either condition so cake events don39t matter 0 ls only history effect if event impacts study results IE study on suicide celebrity kills self in middle of study could raise awareness of problem I Cannot stop an event from happening lReactivity Effectszl 0 Being reactive like social desirability effect 0 Becomes altemative explanation if participants are reacting to being studied rather than IVtreatment in uences DV 0 Threatens intemal validity with fact that participants know what goal of study is I lHawthome effectl lixperiment on worker productivity tried giving them more light looked at productivity after improved Onegroup pretestposttest design increased light improved again 3rought lighting down would have expected productivity to be lower instead goes UP jorkers increased workload no matter what because they knew they were being observed Need control group to have difference but receive same amount of attention to avoid I lPlacebo effectl Reacting to thinking you39re getting something that will have speci c effect Control for it by actually using it in study 0 Group doing everything else equally with other in study I lDemand characteristics Thing subjects are responding to demand of artificial setting of study subjects think they know hypothesis respond accordingly 0 Know what they want you to say 0 Treat both groups same won39t figure it out 0 So how to removecontrol these threats I Conduct a TRUE experiment RA to proper conditions I Be sure to treat groups equally All groups get equal time attention etc Threats related to pretesting or measures over timel These relate to Onegroup pretestposttest design within Preexperiments Testing effect sensitization 0 Pretest sensitizes subjects on topic affects answers on posttest IE smokers taking pretest about habits changes later response lMaturation 0 Comes from idea that people naturally change mature over time I IE reading program checks improvement of reading scores but over course of year kids will improve anyway 0 Natural uctuation is same Seasonal changes also a type of maturation lStatistical regression to the mean 0 Phenomenon that happens because of laws of chance but only applies when starting with extreme scores I IE fail SATs without any changes will improve next time around closer to mean than first time nowhere to go but up I Also occurs when starting high up nowhere to go but down statistical chance you will do worse 0 Chance that you will become closer to the mean whether increasing or decreasing 0 Only real problem if limiting sample to those with extreme scores llnstrumentationl 0 Changed how you were measuring something from timel to time2 IE calculate sales of type of shoe then check sales for whole department I Change guidelines de nitions when they should be consistent lMortality attritionl 0 From timel to time2 some people will drop out Those who drop may be different than those who stay IE people seemed smilier on second round bc all the grumpy people dropped out How are these fixed by true experiments 0 Need to account for what is happening to all groups Lecture 52014 Recap True experiment 0 Manipulationcontrol random assignment intemal Validity If NOT a true experiment or if do exp Improperly then gt 0 Altemative Variables become possible explanations How to remoVe control pretesting threats 0 Conduct a TRUE experiment I RA to proper conditions Be sure to treat groups equally 0 All groups get equal time attention etc Threats to intemal Validity cont Experimenter39s EffectBias 0 Experimenter39s behavior or attitudes rather than treatment IV in uences DV I IE One group told they are handling smart rats for an experiment others told they are handling dumb rats but all rats are actually just randomly assigned then compared I Smart rats went faster in maze because experimenters treated them differently 0 How to control exp39er effects I Same thing again true exp etc but also Automate or script the experiment 0 Read uniformly possibly prerecorded or have ignorant exp39er 0 Only know the most basic instructions IE hand people a packet or have blind exp39er 0 The know there are 2 groups but do not know who was in which 0 Double blind is best What is being done in True Experiments that rule all these problems out Selection bias Use RA History effect Even if a change happens from timel to time2 it is happening equally to both groups washes out as explanation Exp39er effects Must make effort to make sure everything is uniform When studying test actual Variables in these scenarios IE different than smiling example Example study music and leaming R R RQ Does listening to music while studying hinder or enhance leaming 0 Causal needs experiment to answer Possible experiment 0 IV Listening to music DV leaming X Music 01 test score groupl M65 No music 01 test score group2 M78 Posttest only design 0 If data is those means can conclude Music hinders the group39s leaming cannot generalize 0 All study types examined have only 1 IV 1 DV What if we want to test for effects of ANOTHER IV lFactorial Designs I urpose Factors are IV39s 0 To examine the effects of 2 or more IV39s simultaneously 0 Each factor has at least 2 levels conditions I Example DV learning test score Music factor While studying music no music AND Caffeine factor While studying caffeine no caffeine MUSIC NO MUSIC CAFFEINE NO CAFFEINE This is a 2X2 design 2 levels of Music X 2 levels of Caffeine easiest factorial design possible 0 Can have more POP MUSIC CLASSICAL MUSIC NO MUSIC CAFFEINE NO CAFFEINE This is a 3X2 design 3 levels of Music X 2 levels of Caffeine 0 Still only 2 factors Music and caffeine I What if more than 2 factors Music factor pop classical none Caffeine factor caff no caff Gender male female I Now 3X2X2 design Would look like 3x2 for men 3X2 for women with boxes Factorial designs test for 0 Main effects 0 Interaction effects Lecture 52214 lFactorial designs contl Main effects 0 The effect of one IV individually on the DV I Simplest I IE for the 2 music X 2 caffeine study Main effect for caffeine 0 Lower scores w caff than Wo caff worsens leaming 0 Could get as many main effects as IV39s 2 I To test for main effects compare marginal means of DV for each factorIV DV leaming test score Music factor While studying music no music AND Caffeine factor While studying caffeine no caffeine MUSIC NO MUSIC CAFFEINE lM 5 0 M 60 No CAFFEINE Marginal mean is combined mean for a factor 0 Need a mean for people who do caffeine do both averages 0 55 for caffeine group 65 for no caffeine groupthink I If there is a difference there is a main effect I Main effect for caffeine Greater leaming without caffeine than with or caffeine worsened test scores 0 For music match Vertically MUSIC No MUSIC CAFFEINE M50 lM60 No CAFFEINE M 70 lM60 0 Both means for music are 60 0 No main effect for music I Studying with or without music made no difference But gt Main effects do not tell the whole story Interaction effects 0 The unique effect of the combination of IV39s 0 Different effects depending on different combinations I The effect of one IV depends on the levels of the other lVs Examples for a Music X Caffeine interaction Caffeine reduces leaming only when combined with listening to music without music it has no effect 0 To test for an interaction effect graph the l means MUSIC NO MUSIC CAFFEINE M50 M60 NO CAFFEINE M 70 M60 Test scores on Yaxis 100 90 80 70 No caffeine 601 j 50 lt On Xaxis Put one IVfactor Music No Music There is an interaction effect if the lines are not parallel 0 So although caffeine lowered scores overall the effect was worse when combined with music Music actually improved scores when without caffeine For those without music caffeine did not matter With different data MUSIC NO MUSIC CAFFEINE M 90 M 70 80 NO CAFFEINE M 70 M 5 0 60 80 60 Test scores on Yaxis 1o0 9o 80 7o 60 5o O No caffeine linel lt On Xaxis Put one IVfactor Music No Music Main effect for caffeine main effect for music Lines are parallel no interaction If lines cross interaction A word about factors lV39s 0 In one design can have as lV39sfactors I Manipulated Variables EX Music exposure caffeine I Subject Variables lix gender personality traits TV use hilo 0 Can only make causal conclusions about manip39d lV39s not Ss vars O O If Q manip39d vars at all then it39s not an exp39t it39s a survey w factorialtype setup Factorial designs cont ls IV manipulated RA If yes a true experiment If no a survey l QuasiExperiments O O O O O Closer to true experiments than preexperiments Not true experiments no RA but have decent comparison groups Nonequivalent Control Group Design Pretestposttest with quasiequivalent groups 01 X 02 grpl O1 O2 grp2 Use pretest scores to match groups before manipulation Simulating random assignment IE went to 8am to do experimental condition then went to 10am to do control condition If both groups are same can make a case for data Time Series Designs Track many observations over time before and after a manipulation Able to see trend over all as opposed to l pretest 1 post test Same group over and over Singlegroup interrupted time series design 01 O2 O3 04 X O5 O6 O7 08 grpl jlmproves upon the onegroup pretestposttest Example Crime prevention program Politician nds Crimes reported Jan x Feb Actual data Crimes reported Jan x Feb I Solves some threats to intemal Validity testing maturation I Variation take treatment away and measure again Could also compare 2 cities 1 w rec cen l Wo Lecture canceled 52714 Lecture 52915 Research paper still due next Tuesday this way can have time to prepare for final Section tomorrow all about paper get a lot written tonight to get help LPE studytutor session next week Wed Jun 4 69pm in the Hub GO TO THIS We are skipping WithinSubjects Designs both in lecture and book Experimental Research cont one final issue Laboratory Vs Field experiments 0 Laboratory Experiments I Bring Ss subjects into highly controlled setting can be anything as long as it is controlled I High control gt high intemal validity I Arti cial setting gt low external validity I Must watch more carefully for exp39er and reactivity effects 0 Researchers themselves can treat subjects differently must treat them same 0 Reactivity is when subjects are aware something is being done to them I Field Experiments lvlanipulate lVs in the real world ltill an experiment need RA should not confuse with field research in general refers to qualtitative methods lix Littering studies 0 Interested in behavioral contagion to study common transgressions like cutting in line littering etc believe they can get away with it 0 To test iers placed on cars in parking lot randomly assigned people coming to parking lot in nontreatment or control group 0 wherein a confederate would be seen by control group subjects throwing ier out of car I The group that saw the confederate litter also had high littering Ex She Said No TV movie study social awareness type movie 0 Managed to use representative sampling in true experiment 0 Randomdigit dialed Americans who were randomly assigned to 2 conditions I Told to tum to movie channel in control other group told to watch whatever 0 Able to test subjects from within their home 0 Ultimately movie oversimplified rape issue backfired in older male subjects not supporting cause I Very rare and expensive experiment More natural settingbehavior gt higher external validity 0 Increasing external validity in terms of realism Less reactivity IE litter people weren39t even aware they were subjects Harder to maintain experimental control Matter of goals and feasability Content Analysis Quantitative Content analysis is by nature through numbers quantative are identified as such Systematically quantitatively examining the content of communication 0 Often studies of media messages 0 Sometimes don39t even have subjects unless recording and analyzing conversations Used to Describe how muchwhat kind of certain messages there are eg sex on TV types of tweets 0 IE magazine intemet ads and the ideal of women39s beauty prosocial behavior in animated disney lms motivational counseling call targeting obesityrelated behaviors among postpartum women concussion related traffic on Twitter Assess image of particular groups in media e g stereotypes of race gender age political party etc 0 IE rare portrayal of wealthy in media as good race in primetime ads coverage of the orida panther Compare media content to real world 0 Implication that media distorts reality so use reallife statistics to examine I IE gender race and age in video games compared to the US population I No one can say what right amount is but can use stats like population as reference point 0 Examine message changes over time I IE motherhood and sexuality in magazine articles over 20 years 0 Provide background for research on media effects Also a method for codinganalyzing openended data in surveysexp39s Entire studies can be just content analysis Important Issues 0 Sampling I De ne population of interest Ex primetime TV shows FB discussions I Identify unit of analysis for coding Ex number of sex scenes on I V I For TV shows code each episode Scene Each character I For FB pages code each entry Thread 0 Select representative sample ideally I Findings cannot make broad conclusions without are less meaningful without 0 Coding Transforming content into numerical categories I Conceptualize categories What is meant by stereotype portrayal sexism Violence Manifest content Visible surface content 0 IE can see an act of Violence know an implication is of sex naked under sheets 0 Have to be able to define but you can still clearly see something IE couple actually in embrace Latent content underlying meaning 0 IE theme in time travel movies based on correcting past Or that we should not mess with science 0 IE alienation as theme with Disney heroes Belle disconnects with others around her count the movies where this is theme or incidents where this theme is being supported 0 Still need definition but much harder to discem Still a quantitative method which means it wants to be as objective as possible I Operationalize categories IE difference between sexual kiss or familial kiss Distinction between being polite and being submissive I Establish reliability Can take multiples attempts Limitations 0 Purely descriptive I Cannot explain why the content is that way I Cannot conclude anything about effects of the messages I IE news media relying on Twitter is not representative of users or what people take from it 0 Very reductionistic I Idea that science reduces complex topic to a few key variables I Particularly noticeable when coding 1 thing in entire movie Reduces content to codeable concepts only Lecture 6314 Mon Jun 9 special office hours 3430 ssms 4105 TA QampA Wed Jun 11 45 ssms 1009 Qualitative Research Methods Qualitative types of message analysis Contrast with content analysis Analysis of content in nonnumerical approach Subjectively analyze comm messages e g media content conversations Researcher is allowed to give scholarly opinion not unlike pro film critic 0 Rhetorical criticism I Critigue form content imagery delivery of speechespop culture I Without counting points out interesting terms or concepts usually starts with famous examples like MLK Hitler etc I EX use of metaphors in a presidential speech Content analysis can look at same stuff but must count it totally different Discusses that there may be a dilemma for listeners or that it is not very persuasive I Ex themes in students drinking stories on FB I Goal greater understanding andor appreciation similar to literary criticism Get to have an interpretation of its importance or what the speech is trying to do llritical theory aka cultural studies 0 Criticizing something39s role in society 0 Craft arguments about the cultural implicationsoppression esp of gender race class etc of media I IE hiphop dreamworlds39 documentary Marxist or feminist analysis of advertising images putting spin on interpretation I No objectivity because these are positions and ideologies 0 Goal socialpolitical awareness amp social change Qualitative Studies of People other labels often used are interpretive ethnographic or eld research contrast with surveys and experiments Goal to develop rich understanding of peoples subjective experience Desire to see from others perspective can ask what someone is going through Some important features 0 Natural setting I Must be there unlike other methods 0 Researcher is not separate from participants researcher subjectivity I Can even become involved in advocating for activism 0 The subjects guide what is studied I Must be willing to focus on what subjects state is important 0 Inductive theorybuilding I Qualitative use inductive theorizing gt start with observation here Participant observation Participation while being with subjects you39re researching 0 Researcher participates to varying degrees in the events groups under study 0 Can be within the action or attempt to be ignored while there levels vary 0 Natives may or may not be aware of being studied Important issues 0 Typically purposive types of sampling case studies common I Seeking out certain people of interest IE looking for successful CEOs famous ComiCon attendees 0 Construction of detailed field notes amp records I Must happen in moment 0 Finished when achieve saturation I Like in chem when a solution will not hold anymore because so much is added sugar falls to bottom of tea after so much is added I More data will not add any new insight QualitativeField Interviewing llnstructured or semistructured 0 Openended Qs free to change 0 Getting de13 th is key I Trust necessary iypes of interviews 0 Ethnographic conversations I Naturally occurring conversations that can be provoked into more thorough interview I Catching as they happen 0 Indepth interview I Require lots of time Focus Group Groups discuss an issue in presence of moderator 0 About 515 people in group interview 0 Again openended questions 0 Leader should facilitate not control 0 Popular technique in marketing and political research The Trustworthiness of Data Qualitative research is NOT concerned with 0 Reliability and validity of measurement 0 Intemal and extemal validity Instead focus is on quality of researcher interpretations 0 IE picking right examples from FB as opposed to cherrypicking for agenda 0 Should be credible trackable wellreasoned 0 Good to triangulate qual methods where multiple methods are used e g participant observations with depth interviews I Want richness of understanding but must be able to live with messiness Lecture 6514 Mullin special of ce Mon Jun 9 3430pm ssms 4105 TA qampa Jun 11 45pm ssms 1009 Parscore form Final lectures since midterm chapters 3 7 8 9 10 ll 12 not cumulative per se but things that reoccur like validity reliability etc study in chunks to keep on topic IE survey stuff experiment stuff create examples separate from notes Research Ethics Treatment of Subjects Institutional Review Board Must submit research proposal for approval from university39s IRB 0 At UCSB Human Subjects Committee 0 Must complete Human Subjects training Guidelines for Using Human Subjects 0 Participation must be voluntary 0 Must obtain informed consent I Explain to participants lurpose and procedures but not hypotheses lossible risks and discomforts lAbility to withdraw Should protect subjects from harm 0 Should not diminish selfworth or cause stress anxiety or embarrassment Should preserve right to privacy 0 Two ways I Anonymity no way to identify responders any FtF research cannot have I Confidentiality I Should avoid deception 3oth types 0 Outright deception Deliberately providing false information 0 Concealment Withholding key information Any deception must be justified by compelling scientific concems Subjects must be adequately debriefed Research ethics Treatment of Data and Research Reporting Ethical Handling of Data 0 Do not manipulate fudge alter etc data 0 Keep all data in case researcher is accused of falsi cation can prove what was done I Allow other scientists access leporting Findings 0 Support proper peer review 0 Avoid con icts of interest I Funding sources that expect certain ndings Good to be skeptical Consensus is not necessarily a good thing What happens to research after it is conducted 0 Academic publication 0 Practical applications I Marketing consulting policy 0 Media coverage of hot topics I J oumalists AND consumers often lack understanding 0 The File Drawer TA Office hours 130330 Ask to show us how to tell a main difference is occurring Underline key words on test like RA cause leads to for factorial designs draw graph to check for effects and DV needs to be on Y axis Prof QA Quasi at least tries to overcome intemal validity threats and be more true but lack RA In pretests Static group means that groups were already existing no RA not true Field experiment is quantitative true RA Field research is purely qualitative Statistical regression states that by luck your scores got closer to mean or improved Think of examples and use variables to leam how experiments happen mix them up and then try to match them Ta QnA Saturation in Field Research is the repeated concepts that keep coming up Triangulation always better to have multiple methods IE paper survey and interview Factorial designs looking for interaction of 2 Ivs combining to create a stronger effect or main effect of l of the Ivs having an effect when graphing need DV on y axis vertical Whether they are true or not depends on random assignment Cross lagged panel designs in a longitudinal survey helps to get closer to causality bc you are looking at one group39s changes over a period of time X1 is compared to Y2 C2 compared to Y1 Quasi experiments have SOMEWHAT of comparison of matched groups already existing IE different classes Pre experiments are not at all attempting to be true experiments
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'