Psych 7 Complete Course Notes
Psych 7 Complete Course Notes
Popular in Course
verified elite notetaker
Popular in Psychlogy
This 59 page Reader was uploaded by Kaley jorgensen on Monday April 14, 2014. The Reader belongs to a course at University of California Santa Barbara taught by a professor in Fall. Since its upload, it has received 94 views.
Reviews for Psych 7 Complete Course Notes
Eugh...this class is soo hard! I'm so glad that you'll be posting notes for this class
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: 04/14/14
41414 616 PM What is the Scientific method Scientific method guides research methods in psych Testing hypothesis in ways that can be checked scientific thinking science is 1 empirical based on obs 2 objective obs And concl Are clearly defined in ways that allow others to get the same results 3 systematic obs Are organized in a way that provides clear evidence for specific claims what is causing what gt causality 4 based on theory optional EMPIRICAL Science is based on observations vs authority Galieo going against the authority of the church with his theory drawing concl Based on obs Vs authority finally the church accepted it prohibition against arguing from authority in science alters the nature of expertise an expert is not someone who has the correct answer only someone who knows the evidence very well and can make a persuasive argument any consumer of research still needs to evaluate the evidence being presented IIIIIIIIIII OBJECTIVE Ex auras and objectivity some ppl claim to see colorful auras around ppl based on these quotobservations but not available to everyone not objective not verifiable systematic tests measuring devices blind tests Ex therapeutic touch ppl have invisible energy fields and trained professional is able to sense the energy and smooth it out and that will make the person feel better this is proven unjustified so why still do it Placebo affect SYSTE MATI C Ex systematic thought bloodletting get rid of bad blood so you can heal kind of like an oil change replace old blood with healthy blood how did this go on for centuries observations gt did they die couldn39t draw clear inferences about causality drawing poor conclusions because of poor organization how to make this a better credible experiment control group random assignment Ex systematic evidence for therapeutic effectiveness say you work for a clinic that treats anxiety disorders in children most kids seem to get better overtime is this good enough evidence that the ppl are getting better that the therapy is working Is it any different than bloodletting m what is the evidence that therapies help patients Randomized experiments strongest evidence but not always done Question of proper research design ins crucial to evaluation f whether therapy works the way we organize experiments determines the confidence when drawing concl Brief history of randomized experiments randomized experiment more specific than broader term experiment in everyday language common use of randomized experiments began in the 192039s in agricultural studies only sporadic use before then james Iind in 1747 used treatment and control groups to test effects of citrus fruit on prevention of scurvy in 1936 R A Fisher developed stats to analyze results from experiments take a field and divide it into plots and flip a coin to decide what treatment it gets nitrogen or no nitrogen determined randomly what is the problem with this 0 Weather rainfall Intuition as evidence it can be wrong Our perceptual and reasoning mechanisms may not be accurate even when intuition is correct it is NOT scientific evidence usually subjective usually unsystematic change blindness and the scientific method the observation ppl are not good at detecting changes in the environment the hypothesis ppl will be slower to dtect change in two scenes if some visual event briefly intervenes between those two scenes method ecp Using the flicker paradigm two groups one group sees grey screen and one group has no grey screen the result slower with the grey screen Ex hot hand in basketball 30 common belief that basketball players get hot hands in which they are likely to make multiple shots in a row Lecture 114 Describing Behavior The hypothesis ppl have a rightward bias in headturning when kissing The method certain things had to happen for it to work lip contact face to face the result 80124 kissing couples turn to the right many other things have this 21 ratio 0 using right over left TERMS hypothesis statement about the world that MAY be true most ppl turn to right when kissing most state that 2 or more variables are related to each other variable any event situation or behavior that has at least 2 values operational definition of variables a definition of a variable in terms of how it is measured aggression stress sef esteem all variables must have operational definitions to meet the objectivity criterion problems with the experiment the word obvious is subjective it varies from person to person who is I how to solve this problem need ppl who are blind to the hypothesis film the instances and have ppl watch them and determind which side they turn take all subjectivity out of the measurements try to find some correlation between two things do our measures correlate with things that we expect them to correlate with Predictions concern the outcomes of specific studies Hypothesis are usually stated at the level of theoretical constructs Theories generate and explain sets of hypotheses Theory is a conceptual framework that organizes and explains a larger body of facts Theory 9 hypothesis 9 predictions 9 dataobs 9 conclusions EX Theory extraverts have low baseline levels of physiological arousal Relationship between arousal and performance 0 Relationship between mood and arousal 0 At rest extraverts have lower levels of arousal and introverts have higher levels of arousal 0 Extravert needs stronger stimulation to reach a higher level 0 Might prefer a rock concert where the introvert might just want to read to get that same feeling Hypothesis extraverts will be high in sensation seeking Predictions ppl who score highly on the NEO extraversion scale will report stronger desire to try sky diving Datazobs date from specific sample do or do not support the prediction Conclusions theory is provisionally supported or hypothesistheory may require modification 0 Try different things Goals of physiological research to describe behavior to predict behavior to determind cause of behavior to explain behavior predicting behavior beh At voting booth 0 relationship correlation btwn Two variables what o do you vote for dem or repub take diff subjects and plot obs A positive linear relationship 030 look at pp s income and how they vote 030 obs a negative linear relationship observe pp s body weight and observe no relationship obs A non linear relationship 030 education level 030 the direction is changing 030 aka Non monotomic predicting behavior 30 showing a relationship correlation between 2 variables does not show cause and effect 30 problems making causal statements 0 direction of case and effect instead of a causing b maybe b caused a confounding of lurking variables 0 does coke cause polio 0 Correlation gives us ability to predict but 0 Do we know that coke causes polio 0 Maybe temperature is why we have this correlation Ppl drink more coke when it is hot and polio is also more common in higher temperature 0 Temperature confounding variable confounding variable third variable Explains why these things are correlated even though there is no causal effect Reverse causation o Ppl with polio drink coke Ex Correlation between number of drownings in and the amt of ice creams sold there what could be an explanation temperature a common lurking variable more ppl go in the water more likely to drown what about cramps No ice cream caused the cramps 9 a clarifying explanation NOT alternatives Just explaining how one variable caused another much applied research presented in the media is correlational and thus potentially suffers from lurking variable problems EX students whose parents monitor their homework gets good grades example relationship between coffee consumption and risk of heart disease Lecture 116 Introduction to Experiments Be clear on examples about alternative explanations cell phone example correlational research open to alternative explanations does the negative correlation between cell usage and sperm counts at least provide evidence that cell phone usage may cause lower cell phone counts we know they are correlated but we don39t know there could be a causal relationship 9 see a relationship DOES provide evidence causal evidence from correlational date may depend on converging sources of evidence is there a theory that explains this relationship In vitro studies of effects f electromagnetic waves on sperm Controlled experiments in animals to show this effect Although correlation does not directly prove causation it can be useful evidence for causation Correlational studies to not directly prove relationships but it does provide evidence EX smoking causes cancer number of cigarettes and cancer case by case basis and look for converging lines of evidence that suggests if it is causal or not experimental research the researcher manipulates at least one variable and ten measures at least one outcome need randomization One group experiments one thing and the other does something else Ex subjects could be randomly assigned to either 1 drink 5 cups of coffee a day or 2 Drink 0 cups of coffee Then compare cholesterol of the 2 groups after a certain length of time 6 months 1 year manipulated variables are called independent variables and outcome variables are called dependent variables in non experimental research independent are those hypothesized a key Idea in experiments is that as much as possible ONLY the manipulated valuables differ between conditions goal to isolate the independent variable so that39s the only possible cause of diff between conditions in a correlational study individual diff between participants could be a lurking variable coffee drinkers might exercise less in an exp random assignment should hold constant ind Differences that could be lurking variables if sample is large enough random assignment 100 subjects for the coffee exp 0 who exercise and 50 who do not how many exercisers do we expect in each condition How many non exercisers o Expect 25 in each group Is exercise a lurking variable in this 0 NO 0 Exp as sample size gets larger we expect diff to be o On average shouldn39t affect outcome what is true of exercise should be true for infinite set of variables say we have 6 subjects for the exp 3 who exercise and 3 who don39t 0 randomization wouldn39t work in this small of sample say we assign 50 students in the front of the class to the coffee condition and 50 in the back to the no coffee condition need to be random as possible in our coffee ex one group drinks 5 cups of coffee per day and the other group 0 say the shop providing the coffee for the exp Is on a hill Is that a problem 0 Walking up hill is exersice o Manipulating two things coffee consumption and exercise If we get a diff between the groups how do we know if it39s coffee or exersice o Is coffee consumption the only thing being manipulated Experimental control Variables that vary along with the independent variable in an ep that we did NOT intend to manipulate are confounding variables 0 Exercise is reintroduced as a lurking variable 0 Confounding variables co vary with one exp Group but not the other 0 Independent variable no longer exists if there are confounding variables 030 a key ida in exp As much as possible ONLY the manipulated variables differ between conditions 0 sham surgeries in animal experiments 0 trauma of surgery could affect the outcome or amygdala being damaged 0 need to eliminate trauma as a confounding variable solutions to the confounding variable problem in the coffee exp 0 Both groups have to go to the same place 0 Put caffeine in some and decaf in the other don39t tell them which group they39re in o That helps isolate one variable Assessing experiments 030 was randomization adequate 030 Was the sample size large enough 030 Did the manipulation introduce confounding variables 0 Did the investigators manipulate variables they did not intend ti o If so then the alternative explanations exist for differences between conditions In a perfect experiment any diff between exp Groups on the dependent variable must be caused by the manipulated variable 0 Since only the manipulated variable differs between the groups Anytime evaluating research must look at alternative explanations Why does touch therapy produce results for infants Possible explanations babies feel safe and more secure determining that C casues B does not necessarily demonstrate why the phenomenon theories function to supply candidate explinations applied research find solutions to practical problems baby touching helps us solve problems basic research brain mechanisms how they work lots of Why questions lots of research has both of these Lecture 121 Research Ethics Summary for last week Non experimenta vs experimental methods 30 non experimental obs correlational 0 measure variables but do not manipulate them 0 advantages can observe ppl in real life situations can sample large numbers of ppl rep surveys o disadvantage difficult to demonstrate causeaffect relationships 0 have to look at lurking confounding variables reverse direction of causality Experimental 2 things define experiment 0 1 Manipulate a variable 0 2 Randomly assign to conditions advantages can demonstrate causal relationships 0 can solve luring variable disadvantages laboratory settings are often unrealistic may not generalize to real world 0 usually limited 3 validities construct validity does the operational definition of the variable actually reflect its intended theoretical meaning Smiling when angry bad measure Need to measure correctly Do we have good operational definition internal validity can we conclude from our data that one variable caused another How well does it demonstrate causality external validity can the results of the study be generalized to other subjectscontexts Research Ethics be objective as possible repeatable conclusions based on concrete observations ethics is unavoidably subjective dealing w subjects human subjects approval institutional review board needs 5 members w at least 1 member outside of ucsb costbenefits analysis beneficence informed consent autonomy need consent from people before experiment the use of deception debriefing selection of subjects justice fair selection of subjects and cost and benefits including both women and men in studies integrity of data data fraud safeguards against fraud dealing with subjects the use of deception EX blue eyesbrown eye classroom exp NOT AN APPROVED PSYCH EXP 30 hypothesis being included in a group of ppl can affect our perception of ppl from other groups 30 methods a classroom teacher tells her kids that ppl w blue eyes are better than brown eyes Sends them off to playground 030 the deception the kids take it seriously do not know it is an experiment 030 result kids were bullied Kids wanted to have blue eyes principle of autonomy VERY violated were there benefits that outweighed distress Debatable EX migram s obedience exp Method 030 interested in the conditions in which ppl will obey and order that will harm someone else 030 draw straws to see who will be the learner and who will teach 030 want to look at the punishment on memory 030 need to pair which word goes with the other word the deception 030 the learner is not the subkect The electrical shocks are not really being applied The screams of pain are not real exp is not really about memory what is the dependent variable 0 The o of subjects that obey until the end 23 What are the independent variable 0 The learner and the experimenter Vary the distance of the learner isolate one variable learner distance Learner will be in same room as the subject not the subject needs to do acting to show that he is in pain 030 testing if there is an effect of empathy o dropped down 40 Deception is necessary can39t say this is a study about obedience because it will ruin the exp Need a debriefing process with deception dramatic costs from the subjects large cost to impose on subject for psych Exp explained that there were no shocks at the end and explain that lots of ppl went to the end to make subject feel better make sure there is no lasting harm to subjects 0 gave subjects surveys about how the study effected them EX zimbardo Stanford prison exp Hypoth the brutal behavior of prison guards had to do with their social roles rather than personalities Methods assigned college students to play the role of the prisoner or guard in a stimulation Built mock prison to test the hypothesis Want to see if the role changed the personality 0 Randomly assign who is prisoner and who is guard Assigned few rulesguidelines The deception No deception they knew they were going to be playing a role Not same violation of autonomy here the subjects know exactly what they are getting into Everyone quickly became indulged in their roles and it became bad The result After 6 days the study had to be stopped too much stress on prisoners very controversial and some think it violated too many rights and should have never been done animal research test boundaries of what seems ethical research on relationship between work reated stress and ulcers one monkey of search 4 pairs the executive press lever at least 1 every 20 seconds or they will get shocked 26 shiftsday 7 days a week results 34 executives died of ulcers on days 9 23 and 25 passive monkeys did not develop ulcers ulcers develop during rest periods NOT stress periods integrity of data data fraud data seemed too perfect identical error noise in diff exp Difficulty with replication Marc Hauser Harvard psych primarily studies nonhuman primates cognitive abilities similar to those in humans mirror recognition ability to recognize subtle variations in sound patterns and thought to be a human specialization for language humans pass this around 18 months monkeys pass it as well argued that new monkeys could also recognize but others did not see any evidence at all no one could replicate his results was not credible clues of fraud could not replicate students in lab reported suspicious of wrongdoing to the university triggering a 3 year investigation integrity of data safeguards against fraud a peer review process competition database repositories replication Lecture 123 Measurement Exam 1 on Tuesday Construct Validity is it measuring what its supposed to measure Intelligence operational definitions intelligence is related to the speed of mental processing think creatively variety of mental and physical abilities a general factor 0 IQ Weschler adult intelliqence scale verbal subtest comprehension 0 why should we obey traffic laws and speed limits Information 0 What is entomology o What is the temperature Performance subtest Block design 0 Arrange blocks into target strategically these tests seem to be measuring different things History of the intelligence test Alfred Binet thought you could improve performance if you gave extra attention believed you had to do well on all tests and sum across developed tests and came up with a to express how students were doing IQ Is the IQ test a reliable and valid measure of intelligence RE LIABILTY test retest reliability if an individual is tested at two different times will the scores be the same consistency Measure ppl multiple times and see how well the scores correlate Internal consistency reliability spit haf reliability if you divide the items of the test in half would both halves have the same score 0 Should have high correlation with the two halves of the test PEARSON PRODUCT MOMENT CORRELATION COEFFICIENT r Can range from 100 to 100 IQ generally does not deviate more than 5 over an individual s lifetime subscales of the WAIS correlate high with one another Charles Spearman proposed g as the general intelligence factor that is called upon for any intellectual task and thus explains high correlations between IQ test subscales ppl with more g do better on all scales internal consistency reliability provides evidence for q 0 they correlate with each other measure g provides a general measure of intelligence Performance on any one subscale then a function of both g and skills specific to that scale construct validity face validity measure seems to measure the intended measure criterion oriented validity measure relates to theoretically predicted criteria 1 Predictive validity 2 Concurrent validity 3 Convergent validity 0 4 Discriminant validity take a measure and correlate it with other stuff and see how it relates good grades and no criminal record VALIDITY Converqent validity Is it related in a predictable way to other variables that are supposed to measure the same thing Correlate one test with another test where we think they should correlate Predictive validity Is it related in a predictable way to future behavior 0 IQ tests correlate 50 above with grades in school Concurrent validity Does the measure correlate with current behavior Do different groups of ppl differ on the measure in expected ways Discriminant validity Does the measure assess the intended construct better tan other constructs Measurement Scales Nominal scale a categorical variable no numerical or quantitative properties Ex in the 1989 Buss study male or female subject Ordinal scales a rank order of the levels of a variable The intervals between numbers are not necessarily o The list of your favorite songs Interval scales the intervals between the s are in size and there is no absolute 0 No true 0 Can39t form ratios Ratio scale Intervals between s are also in size but have an absolute 0 0 Measures such as length weight time 0 Someone can be twice as tall as someone else Theory 9 hypothesis 9 predictions 9 dataobs 9 concl Evolutionary Psychology theories about other theories and the organization of the brain Metatheory Human brain contains a collection of specialized processing mechanisms designed by natural selection to address specific problems encountered by our ancestors over the course of human evolution Examples of adaptive problems Visual scene analysis Mate choiceattraction Navigation Food choice Predator avoidance Parenting ALTERNATIVE The brain has general learning mechanisms OOOOOO EP specialized mechanisms 9 mating parental investment theory PARENTAL INVESTMENT THEORY PREDICTED THAT THE GREATER TYPICAL PARETAL INVESTMENT IN OFFSPRING BECOMES A LIMITING RESOURCE FOR THE REPRODUTCIVE SUCCESS OF THE OPPOSITE SEX females of most species make larger typical parental investment than do males invest for 9 months and cannot make offspring during that time investment of time and energy to produce a baby once kids are born they need lactation which turns off ovulation seems to be a 3 year period until women can conceive again a man has a small investment of time 0 men can reproduce at a faster rate than women IF they can gain access to women Bateman s data with fruit flies took 3 male and 3 female flies and put them so they could mate took some with white eyes and red eyes so he could identify who was who39s offspring can compute how many eggs the male and female had eagerness for sexsexual variety Coolidge effect 0 m Clark and Hatfield study Lecture 130 Naturalistic observations goal get accurate picture of behaviors events settings and just watch them analyze observations and form hypothesis Q05 hoc need to operationally define variables presence of red females can be a source of hypothesistheory participation vs on participationconcealment vs non concealment problem of reactivity behavior changes as a function of observing ppl know they are being watched so they act different solutions to reactivity concealment so they don39t know you are there habituation hang out there long enough so they get used to you important issue of how to categorize events wo operational definitions how do you ensure obs Are objective Robert Ciadini s studies of persuasion watched car salesmen and saw what techniques they use got a job as a car salesman low ball technique customer given price much lower than salesman intends to sell car in order to induce customer to agree to purchase intuitive concl Based on naturalistic obs Low ball lab study Subjects were undergrads called to participate in a study Iv fill disclosure 7am study vs commit then disclose DV behavioral compliance Result 24 full disclosure vs 53 commit first Systematic observations Goal careful observation of one or more behaviors in a particular setting 3 things make it systematic interested in only specific behaviors quantitative compute stats hypothesis driven nature quantitative methods components amp issues coding system how to measure behaviors need recording equipment reliability interrater reliability of coding sampling what segment of time to record limitations reactivity all these careful methods used to record and code the behavior can also affect the behavior results may depend on the coding system and how the behavior is operationalized diddlinq in baboons fondling in genteelly is a particular intimate and risky interaction considering that male baboons have high potential for aggression and that hitting or biting males gentalia costly siqnalinq theory signals can be regarded as more reliable and honest if they are more costly expensive and hard to fake diamond engagement rings why Cost lots of money and it serves no purpose why not get a sports car Has a purpose why go to college Costly signal that says you will be a good employee toearn Hard to test that ppl learn stuff Economic value of college degrees probably follows from their status as costly signals Systematic observational study in Baboons ypothesis intense greetings such as genital diddling are costly signals of close alliances prediction positive correlation between amount of time spent togetheramt of grooming and the frequency of intense greetings coding system time spent together amount of time within 15 meters of each othergrooming is defined as one animal manipulated fur or skin of another sampling look at 1 male for 30 minutes and rec behavior recording equipment obs 30 software in computer reliability not reported reactivity animals are on display in a zoo and are habituated to human observation results positive correlation case studies goal to provide detailed descriptions of behavior of an individual usually in rare circumstances most case studies are qualitative vs quantitative limitations conclusions drawn are often post hoc difficult to determine causality examples psychobiography patients patient studies goal to link brain activity to behavior modularity of function paul broca broca s area language production could still comprehend speech but not say anything 9 language is somewhat modular organized 0 can knock out some parts of it but not others Archival research Nature quantitative Methods compoents amp issues Mining data 3 types Statistical records survey archives written records Content analysis devise coding system that raters can use to quantify the data Limitations Info can be hard to find and cant be sure of accuracy Archival research ex homicide data Martin daly and Wilson set out to test the hypothesis regarding patterns of homicide Used archival data from police departments and other dov Records Reasoned that homicides are objective and serious enough that records will be kept Ex hypothesis ppl will be less likely to kill their genetic kin than other classes of individuals Lecture 24 SURVEY RESEARCH keep it simple amp avoid technical terms easy to interpret make sure the question is directly relevant to the hypothesis THINGS TO AVOID Double barreled s amp loaded or leading s EX Q how many times a week do you M avoid loading or leading questions suggests the answer in the question don39t do this HALO EFFECTS avoid associating a position with a person so that responses may reflect feelings about the person rather than the position do you agree with the president that Hussein is a grave threat to the US if they like the president then that can sway their opinion take out the person when asking the question DOUBLE BARRELED QUESTIONS do you believe that airbags are unsafe and expensive Do you believe they are unsafe Do you believe they are expensive Ask the questions separately sometimes not so obvious they are double barreled ex college students should receive grades in their courses because this prepares them for the competitive world outside of college still two separate things restate this into multiple questions maybe they don39t agree w the reason OPEN ENDED VS CLOSE ENDED Open what is the most important thing for children to prepare them for life Advantages can discover unexpected answers don39t restrict responses Disadvantages takes more time respondents might not think of things that you are looking for Closed give a list of options and you choose one Advantages easier to code make sure the things you are interested in are in the response options Disadvantages might have given other responses if they were given the opportunity to do so what is the most important thing for children to prepare them for life Open 5 said think for themselves Closed 62 said think for themselves Big differences Estimating behavioral frequenciesdurations unless events are rare enough to count subjects use estimation strategies that may be affected by response alternatives cosed ended response questions Make sure the options are mutually exclusive Make sure the options are exhaustive Ex how long will someone wait in line for freebirds when the lines are long on a thurs fri or sat Check one 1 min 5 minetc comparing men to women RESPONSE SCALE SENSITIVITY insensitive response scales may produce ceiling or floor effects ex women shower more than men insensitive 1 once a year or less 2 more than once a year but less than once a month 3 between one month and once a week 4 least once a week problem all answers will be clustered ceiling effect effects of question order ex how satisfied are you w your life overall Answer depends on what info is currently on your find accessible other things could prime you to say how you feel EX subjects asked to list 3 positive things in their life or 3 negative randomly assigned to which groups then asked to rate their life satisfaction ASSIMILATION EFFECT when info currently in mind produces more evaluation of a target Feelings as information subjects are called by phone supposedly from another city for a phone survey called on sunny days or cloudy days in one condition asked how satisfied they are w their lives in other condition asked how the weather was conclusions about the survey cognition ppl construct answers to surveys on the spot using any relevant info they currently have proper interpretation of surveys requires knowledge of context not only obv Mistakes items that surround the question or the order anything that can prime a certain answer response sets tendencies to respond a particular way wo giving accurate answers social desirability bias tendency to respond in any way that makes subject look good how do we ensure we get honest answers EX common finding in survey research that on average men support hetero Partners than women do but it takes two ppl to have sex so they should be social desirability bias working in opposite directions men want to brag about how many partners they have had women underreport the bogus pipeline a fake lie detector test attach subjects to fake wires but they aren39t doing anything found that males stayed the same in their responses but women changed their responses random assignment vs random sampling random assignment to conditions assigning subjects to the conditions of your experiment such that every subject has an chance of being assigned to any condition random sampling from the population choosing subjects for your study ex literary diqest poll predict who was going to win presidential election sent out 10 million surveys to 2 groups of ppl to upper income and lower approx 2 million ppl responded what went wrong weren39t sampling randomly selection bias ppl who were wealthier were overrepresented R sampinq frame actual population of individuals from which a sample is drawn selection bias results when sampling frame is not representative of the population of interest this study had a big selection bias non response bias 25 ppl returned surveys BASIC DEFINITIONS Population group of ppl you are interested in Sample some subset of population Samplinq frame the actual population of individuals or clusters from which a sample is drawn Response rate o of ppl in sample to who complete survey Sampling simple random sampling each member of pop has chance of being included best choice stratified random sampling divided into groups and random sample from each group do this if you want to study and compare groups of interest cluster samplinq clusters of individuals are identified then clusters are randomly selected All individuals win a selected cluster are sampled go to schools and survey the group drug use in LA go to diff schools these top three are probability sampling haphazard convenience samplinq survey whoever is convenient to survey problem don39t know that your group is representative of a larger population quota samplinq uses convenient sampling make sure you have a representative amount of each group these are nonprobability sampling sample size and sampling accuracy a rep random sample can often est the true pop Mean accurately w sample that is small relative to pop mathematical formula for determining confidence interval relies primarily on size convenience samples used in much psych Research good if you are studying relationship between specific variables 0 ex w haphazard or convenient sample can still have internal validity for that sample and still has value 0 issue of external validity haphazard 26 Midterm Explanations Experiments manipulation of a variable random assignment independent variables manipulated dependent the outcome reverse direction of causation does not happen in experiment hold all lurking variables constant 0 so they cant be the cause we run experiments allow you to rule out other possible explanations sample size large groups should be equal smaller group could have things happen by chance one large group gives specific answer if there is random assignment then it is a true experiment it CANNOT be correlational if there is random assignment Lecture 211 experimental design Behavioral endocrinology The study of how hormones affect behavior and vice versa Arnold Berthold in 1849 removed testes of a rooster stopped crowing no more aggressive or sexual behavior reimplanted one testis somewhere in the body but not in its initial spot the reimplanted testis restored these behaviors o resumed crowing etc castration in humans have similar effects alessandro moreschi high pitched voice singer voice didn39t change in puberty castrated males experimental animals showed absence of sexual responses to females injection of testosterone into castrated males resotred sexual behavior to normal levels reverse direction of causality Would sexual behavior or exposure to females increase testosterone 30 Post test only design Can exposure to females increase mae s testosterone the first studies to investigate this used posttest only designs Purvis o Haynes 1974 Cages with 2 compartments separated by wire mesh barrier See lecture for more simplest design exposure to random variables and measure once blood letting ex randomly assign 1000 ppl to have monthly bloodlettings for 6 months vs no blood letting at end of 6 months have doctor blind to condition rate general health post test only design early studies on testosterone responses to females had to use post test only designs because had to sacrifice animals to get enough blood could not measure the dependent variable more than once ex before and after manipulation improved hormone 30 Pretest Postest Design measure the dependent variable before and after can see whether they changed over two groups advantages small sample size can ensure groups are before manipulation pretests lets you get similarities between two groups 0 same base testosterone on the independent variable can measure changes within an individual instead of only across tgroup as a whole experimental manipulation may have different effects on different ppl can assess effect of manipulation in spite of mortality effects detecting changes in individuals set randomly assign to medicine vs control placebo for treatment of some condition using post test only design don39t know what the initial starting point is adding a pretest helps see what the starting point is o easier to detect change adding a pretest how does measuring in beginning and at the end solve the problem of drop outs 0 From their pretests we could see what kinds of ppl dropped out Based on initial test scores if we only have a post test design then the only data we have is ambiguous does blood letting help health or did all of the unhealthy ppl drop out Add pretest 0 Can also detect changes from before to after Disadvantaqes Pretests could effect results and Solomon 4 qroup desiqn tests for effect of pretest Groups 1 posttest only control group 2 posttest only experiment group 3 pretest control group posttest 4 pretest experimental group posttest stress of going through blood draw in pretest wipes out testosterone response and see no effect Its not that there is not effect its that we cant use a pretest because we cant test the variable we are interested in Ex puzzle performance Teaching by Read ppl solve it Watch ppl solve it how would we use a pretest Have them try the puzzle without any learning of how to do it To get a starting point Could have a practice effect 9 ceiling effect practice makes you better and has nothing to do with reading or watching adding a pretest wipes out ability to see effect of manipulation 30 repeated measures desiqn every subject is in every experimental condition how would you do the blood letting exp With this design all ppl undergo 6 months of bloodletting and 6 mo Of no bloodletting dependent measure doctor health ratings at end of each 6 month per Half subjects bleed first half no bleed first Hold everything constant except when blood comes out Higher scores for ppl with no bleeding first Lower scores for ppl with bleeding first Why The ppl from the first blood letting are still recovering The other ppl haven39t experienced blood letting yet so they are in better health The order you go through conditions affects the outcome Advantages Controlling for individual differences in variables Each subject is his or her own control making it easier to detect the effect of manipulation with fewer subjects Why is sample size important in a randomized posttest only experiment 0 More evenly distributed the lurking variables would be 0 Will get more number of ppl 0 Individual differences are Repeated measures design need large sample size 0 All individual difference variables are guaranteed to be the same because they are the same ppl 0 Both experiencing control and exp Group 0 Comparing subjects to themselves 0 Allow us to see effects of manipulation with small sample size Disadvantage Order effects Going through in a certain order could affect outcomes Post test only design 5 per group variablility within each group is not related to manipulation repeated measures design each subject is tested twice N 5 Can still see effect of manipulation because you control for individual difference variables because you are comparing subjects to themselves Variablility within differences is removed bc compare subjects to themselves b4 and after 30 Matched pairs design subjects are matched on a characteristic related to the dependent variable and then randomly assigned to conditions technique to ensure that experimental groups are before being subject to the experimental manipulation usually used with SMALL sample sizes EX You want to test hypothesis that computerized tutorials will improve math performance more than books students 1 hr study per In one or other condition each day for 1 month dependent variable is score on math test at end of month get 16 students with low math grades 8 F 8 M which one works better so they can purchase one 0 Need to test which one is better What is the problem with running a random post test design 0 Small sample size 0 Flipping a coin could still be unbalanced by chance Need to make sure the groups are on IQ before you experiment Researchers have previous evidence that IQ scores relate to performance on 0 Take two girls with highest score and boys and make them pairs 0 Keep going down the list and match them up by IQ and gender 0 Then flip a coin to decide whether they go to book or computer 0 Now we know that our results wont be from other variables Repeated measures advantage since comparing individual to themselves decrease effects of subjects variability and make effects of independent variable easier to detect disadvantage order effects matched pairs advantage similar to RM design without having prob of order effects dec subject variability disadvantage cost of identifying and measuring matching variable procedure is worthless is matching variable is not related to dependent measure best case may be when DV IS the matching variable may be unnecessary with large enough sample size in principle set of potentially relevant matching variables is infinite Lecture 213 order and expectancy effects Repeated measures participants in BOTH conditions could be more than 2 conditions could be 10 could go through each condition more than once can use a smaller sample size because you are comparing subjects to themselves and reducing subject variability disadvantage order effects taking the same test twice is confounded with the type of music 0 two things are covarying 0 you are taking the test twice so you already know what the test is order effect practice get better with testing 1 zitigue get worse with testing contrast how to control for order effects complete counter balancinq present the different conditions in every possible order for 2 conditions controlexperiment only 2 possible orders for 4 conditions 24 possible orders n remember repeated measures use smaller number of ppl order effect revealed by counterbalancing 2 subjects go through in on order and the other half through another order counterbalancinq ensures that order effects affect each experimental condition equally removes confound between order amp condition it does not make order effects go away can see whether there are effects of condition above and beyond any order effects if order effects are very strong strong practice effect may be best not to use repeated measures design might use post test only design instead contrast effect the experience of one condition produces an opposite effect on the experience of a subsequent condition type of carry over effect ex judge the heaviness of 2 weights one lighter than the other on a 110 scale how to control increase time between conditions complete counterbalancing present the different conditions in every possible order Latin square 1 Each condition appears at each ordinal position 2 Each condition precedes and follows each condition one time 0 controls for contrast effect letters experimental groups conditions the rows are the order the numbers does not present every possible order 0 Only shows 4 Controlling for the practice order effect 0 They all benefit equally Since we only have 1 order we cant control for every order effect Expectancy effects as confounding variables In the context of an experiment a confounding variable is a variable that varies along with the independent variable Subjects and experimenters expectations may be variables that vary with the manipulation and this threaten internal validity Subject expectations Demand characteristics features of a study that reveal the hypothesis and the condition that subjects are in Subjects may alter their responses to be consistent with the hypothesis Placebo effects effects of the expectation of having received a treatment Drug trials are usually run with a placebo control groups in which participants receive a pill with no active medication VIDEO he was in the placebo and the medication condition repeated measures which placebo is better Lecture 218 Video Compare placebo effects Which placebo is better NH Grant Sham Device pill placebo or treatment for arm pain Thought sham procedure would be stronger that just the pill Results the fake pill was making ppl better than the fake acupuncture Not safe to assume that these can both be used on ppl because they could just be getting better overtime naturally We don39t know that these are helping more than someone who does nothing at all conditions seem to improve overtime naturally Real vs sham acupuncture is all acupuncture a placebo technique 52 athletes with rotator cuff tendinitis 1 had symptoms rated by an orthopedist before and after real acupuncture did better than sham acu Placebo effects Medication group Treatment beliefs uncertain expectancy Placebo group Treatment beliefs uncertain See much stronger evidence for placebo effects is to lie to subjects and tell them they are receiving treatment even when they are not Cannot lie because of research ethics but do it anyway Ex tell subjects they are drinking alc When they are not and half subjects did not know they were drinking alc 2 independent variables what they are telling the subjects they are going to drink expectancy half are told they will be drinking just juice subjects could not detect which they were drinking results cover story taste rating for different brands of vodka or tonic drink as much as you want to get accurate tasting tested hypothesis the loss of control hypothesis that states a small amt of alcohol triggers loss of control in addicts who then can39t stop drinking more result subjects who EXPECTED vodka drank more than those who didn39t regardless of whether there was any alc In their drinks ex set up situation where subjects could deliver shocks based on learning mistakes the only thing that predicted how much shock was delivered was belief that they were drinking vodka delivered more shock subjects implicitly think they can do things like shock ppl experimenter expectations can introduce bias when the experimenter knows which condition subject is in experimenter may treat groups differently experimenter can record dependent measures differently solutions to experimenter bias make experimenter blind to condition run subjects simultaneously for all conditions so all are treated the same automated recording of dependent measures reaction time experimenters we trained to treat subjects the same way developmental research designs cross sectional research subjects of diff ages are studied at one pt in time goal compare age groups on dependent measures 0 age cohort is independent variable 0 measures like attitudes behaviors test performance etc are dependent variables disadvantaqes age and cohort are confounded cannot differentiate changes from cohort effects cohort is roughly synonymous with generation 0 groups of ppl who were born roughly at the same time 0 have experienced common cultural technological and other environmental influences at the same age war depression tv computers advantages cheaper amp faster collect date on age differences in one day that might take 50 years with longitudinal design common in child development research in which may not expect strong cohort effects in infants across small time spans longitudinal desiqn measure same subjects on dependent variables at diff ages can document within person age related changes that cross sectional designs cannot may reveal cohort effects when compared to results from cross sectional studies Flynn effect across many industrialized nations absolute performance on IQ tests have been increasing rapidly over the last 100 years about 3 points per decade for overall IQ scores about 9 points per decade 18 points per generation for tests of fluid intelligence for one test of fluid intelligence Flynn found that someone who scored in the top 10 100 years ago would score in the bottom 5 today explanations for Flynn effect artifact of sampling measurement true increase attributable to changes in the environment limitations of longitudinal desiqn selective attritionmortality longitudinal studies have a rate of drop out in general healthier and better educated individuals are more likely to stay in sample over time selective attrition makes generalization of age trends more difficult especially problematic in studies of older adults history effect sequential desiqns family of designs that combine cross sectional and longitudinal can test both developmental and cohort effects within same design start with cross sectional and then follow them longitudinally pilot studies trial runs with small of ppl training testing for suspicionreactivity testing strength of manipulation manipulation checks assessing the construct validity of independent variable manipulation are you manipulating the theoretically intended construct testing hypothesis that anxiety hurts memory manipulation check tests whether the experimental condition actually induces anxiety often performed with surveys after the study types of manipulation straightforward usually no deception simply present different types of stimuli across conditions and measure their effects Lecture 225 complex designs Factors and levels factor independent variable 0 experiment randomly assign subjects to drink coffee or drink de caf 0 one factor whether or not they drink coffee levels refer to the different conditions within a factor 2 levels in the coffee example regular coffee vs de caf experiment with a 3rd group drink caffeinated soda would have 3 levels but still one factor one factor designs with gt2 levels 2 basic reasons for having gt2 levels need more than 2 to detect curvilinear or complex relationships between variables may want multiple control groupsbe interested in multiple conditions bandura study of imitative aggression children 35 yrs assigned to 1 of 3 conditions observe aggressive adult beat up bobo doll observe nonaggressive adult not beat up bobo doll no exposure to any adult model matched pairs desiqn children were enrolled in Stanford universitys nursery school rated by teachers for degree of aggressiveness matched for aggressiveness before being randomly assigned to conditions hypothesis 1 kids exposed to aggressive models would perform more aggression than those in the other 2 groups 2 exposure to nonaggressive model would inhibit aggression prediction no exposure kids gt aggression than nonaggressive condition dependent variable measured in different room in the absence of the adult models prior to the test imitation however all subjects were subjected to mild aggression arousal to unsure that they were under some degree of instigation to aggression experimenter brought kids into room with attractive toys and after letting them get engaged with them 2 min remarked anyone play with them and that she had decided to reserve these toys for the other children modeling apparently doesn39t work without this added instigation observers counted of aggressive acts in specific categories behind one way mirror over 20 minute interval results 0 the aggressive condition was way higher than the nonaggressive or control for aggressive acts factorial designs factorial designs involve more than 1 independent variable factor all levels of each independent variable are combined with all levels of the other independent variables ex balanced placebo design study factor 1 alcohol expectations level 1 expect vodka level 2 expect tonic factor 2 alcohol intake level 1 drank vodka level 2 drank tonic multiplicative notation 3x4 factorial design 2 factors with 3 and 4 levels 9 12 cells total 3x4 2x2x2 facrotial design 0 how many factors 0 How many cells 8 Usually averages are in the cells Main effects Are the effects of each independent variable considered separately Ex does actually drinking vodka produce higher levels of shock delivered to a learned than drinking tonic main effect of alcohol intake Main effects compare the effects of levels within a factor averaged across all levels of other factors The of possible main effects is to the of factors Ex factorial design Subjects randomly assigned to a 2x2 factorial design DV is performance on math problems correctly finished out of 100 in specific time interval Factor 1 testing condition alone or in front of an audience Factor 2 difficulty of problems easy or hard Post test only design w random assignment across 4 cells A main effect looks at the effect of 1 factor while ignoring the others Interaction effects When the effect of 1 independent variable depends on the level of another independent variable Cannot fully understand influence of one independent variable wo reference to another independent variable Cannot fully understand effect of audience vs alone wo reference to difficulty factor 0 If easy problems better to have an audience 0 If hard problems better to be alone For main effects compute the averages for each condition and plot as bar graphs which is larger For interaction effects plot all of the data as line graphs Nonparallel lines indicate an interaction Simple main effects Simple main effects examine the effects of one factor at one level of another factor Ex do subjects solve more math problems in an audience vs alone when the problems are easy 0 Question is isolated to easy problems one level of the difficulty factor Are extraverts less distracted by noise College students each take 2 different versions of reading comprehension test once with tv on in background and once in sHence Filled out personality test first and selected 10 introverts and 10 extraverts Predicted an interaction effect extraverts would be less distracted than introverts Ex question on factorial designs A researcher is interested in the effects of music and lighting on exam performance In a 2x2 independent groups factorial design subjects are assigned to take the same exam w music being played vs no music being played and in a brightly lit rm Vs dimly lit room Music and bright lights mean 9 Music and dim lights mean 5 No music and bright lights mean 5 No music and dim lights mean 1 Main effects of music 95142 7 51623 YES there is a main effect 7gt3 Main effect of lighting Yes 7gt3 Interaction between music and lighting No plot parallel lines The advantage from music is the same with bright or dim lighting Music does not depend on lighting NO interaction effect Why is the 9 music bright lighting higher than the 1dim lighting and no music Sum together the main effects Lecture 227 Mixed designs Have at least 1 within subjects factor and one between subjects factor In within subjects factors the same individuals appear in all levels of the factor Ex non tactie exposure to females cause male rodents testosterone to increase above baseline Condition 1 Take a pretest to measure testosterone saliva Interact with a WOMAN Then measure testosterone saliva Condition 2 Pretest as above Interact with MAN Then measure test Again We can analyze this as a mixed factorial design What are the two factors Between subjects who the subject talks to woman or a man Diff ppl in diff groups and no one talks w both of them In one condition or the other 2 exp Group man vs woman x2 time of measurement pretest post test hypothesis test Change from baseline with be more in female condition than in male cond How do we translate this hypothesis in terms of main effects and interaction effects is the hypothesis about a main effect or an interaction Interaction effect the effect of pre vs post test will depend on the level of the other factor Hypothesis is about an interaction effect does the change from pre test to posttest depend on the experimental group Is the change from pre to post test more positive in the female vs male Does not confirm hypothesis Test Went up more in male condition than female Not just saying yes or no has to be generated by the right pattern This is the opposite of what we predicted Want to focus in on one Say we want to know whether test Also went up in male condition even if less than in female condition What type of effect 0 Simple main effect isolate the pre vs post to the male factor Repeated measures factorial design driving simulation study with 2 factors 2 drunk vs sober x2tired vs rested Say we spent 20 subjects in each cell in a repeated measures exp How many cells How many subjects 0 4 cells 20 subjects NOT 80 repeated measures 0 each subject is tested 4 times each 0 might wait a week between each test to diminish carryover effect 4 weeks per subject how should we deal with order effects 0 Complete counter balancing NO we cannot because there are only 20 subjects cannot have 24 orders 4 0 Latin square best technique oA10 oB70 oC70 0 D90 40 Tired 80 rested 40 80 is there an interaction effect 60 point diff between tired 20 point diff between rested the combination of drunk and tired being tired multiples being drunk vice versa one factor depends on the other factor only factorial design can see how both interact together independent groups post test groups factorial design now we need 80 subjects diff ppl still 4 cells 3 factor sex of the subject M or F 2x2x2 male vs female x drunk vs sober x tired vs rested factorial design combine all factors make 2 diff 2x2 tables 1 for males 1 for females 3 way interaction Drunk sober tests whether a 2way interaction differs at levels of a 3 factor no 3 way int because the pattern is the same for men and women parallel lines main effects w gt 2 levels say that in math example from last time instad of a 2 alone vs audienct x2 easy vs hard design have a 2 alone vs aud x3 esay vs med Vs hard design 6 total cells and want to assess main and interaction effects if any levels differ form one another than there is an effect interaction effect does audience depend on how hard problems are Yes 35 vs 70 Lecture 34 QuasiExperiments designs that lack all of the control features of true experiments 0 a research design in which an experimental procedure is applied but all extraneous variables are not controlled 0 definitions vary from person to person inability to randomly assign participants to conditions most common reason do not allow the same confidence in causal inferences as true experiments lower internal validity not controlling all extraneous variables can still provide evidence but have to evaluate alternative explanations one group pretest posttest design same as pretest posttest design but no random assignment to multiple conditions no control condition EX hot dogs increase IQ scores take IQ test 9 eat hot dog 9 retake IQ test this is a quasi exp Because of lack of control condition could have had a practice effect nothing to do with hot dog ex empirical study to test the effectiveness of a relationship improvement program marital satisfaction 9 share feelings program 9 marital satisfaction rating same as hotdog design 0 no control group 0 share feelings program may not be a cause problems with one group pretest posttest history effect 0 any event that happens between pre and posttest is confounded w the manipulation 0 ex in the share feelings study a big study reported in media between the pretest and posttest showing that married people live longer Probem maturation systematic changes in ppl that tend to occur overtime are confounded w the manipulation ex study to test effects of vitamins on kids growth height measurements at start and end of 3rd grade w no control group Problem 0 Give them a placebo tell them that they might be getting real pill or not o w random assignment then we can conclude that vitamins work on growth ex drug study to treat symptoms pretestpost symptoms with no control group maturation effect 0 Could be getting better overtime natural changes 0 Cant tell which is causing a change in scores Maturation effect problem to share your feelings program 0 Ppl who are together longer report happier marriage anyways Testing any effects of the pretest on the posttest practice Instrument decay Posttest measurement of dependent variable has changed since pretest measurement observers changed how easily they observe certain behaviors 0 Ex if more motivated to count specific behaviors cigarettes smoked at beginning vs end of study may look like smoked less but really just decay in recording outcome Regression to the mean Problematic when subjects are chosen because they score high or low on some measure Occurs when scores are unusually high or low due to random factors at one testing occasion on next occasion random factors are likely to be absent and scores return to the mean Ex basketball player shoots 40 on average One game shots 90 o This one game was based on change 0 What is player likely to shoot in the next game 40 o long term average regressing to their mean 0 problem for one group prepost if we choose subjects bc they scored low on pretest part of the low scores may be due to chance such that expect improvement on post even wo manipulation 0 therefore scores will go back up because they are regressing to the mean 0 people often misperceive instances of statistical regression belief that punishment improves behavior praise hurts it maybe they are acting out because they are having a bad day not improving from the punishment but regressing to the mean sports illustrated jinx 0 team performance is imperfectly correlated from week to week 0 excellent performance one week is likely to be associated all problems are solved by control group problems are serious enough that they lead us to O Paleolithic diet one group prepost design Most of human history before agriculture humans thought to have consumed very different diet than today 0 Lean meats o No dairy Modern diets in industrialized countries many new components absent until recently Diseases of civilization in wealthy countries like the US high rates of clogged arteries heart disease diabetes etc In modern hunter gatherer societies these diseases are almost absent ppl are more likely to die from viral or bacterial infection Paleolithic diet study Pp s health improved when they were brought into Paleolithic diet Nonequivalent control group design no random assignment but there IS a control group CTE and non control group Chronic traumatic encephalopathy CTE is a degenerative brain disease Post mortem brain examinations find specific markers Symptoms dementia depression abnormal aggression etc Thought to be caused by blunt force traumas to the head A number of ex nfl players diagnosed with CTE after death Hypothesis force of repeated collisions in NFL is causing high rates of CTE in players How can you design a study to test whether playing football causes CTE Randomized exp Precluded by ethics Next best design 0 Matched pairs design compare nfl player to Other athletes 0 Use random assignment Selection effect in some non equivaent control group designs subjects selfselect into experimental group Brain damage patients require nonequivalent controls Ex damage to the medial temporal lobes hippocampus appears to selectively impair the declarative memory system Nonequivalent control group make them similar on every level except the independent variable in this case brain injury Higher internal validity Random exp Lecture 36 single case designsstats review Single case designs Want to test causal effects of a manipulation on a single or small number of subjects Different from case studies because investigator introduces some form of manipulation Employ time series designs ABA design 0 Baseline measures treatment measures baseline measures Limitations of reversal designs Some effects of treatments are difficult to reverse Reversal designs most useful for seeing short term effects of treatments Multiple baseline designs 0 Introduce treatment at different times for different subjects 0 Introduce treatment at different times for different behaviors 0 Introduce treatment at different times for different situations Stats review Many research designs intended to provide evidence that one variable caused another 0 In a true experiment does mean score in one experimental group differ from another group statistical significance assesses the probability that results could be due to chance rather than the hypothesized cause corresponds to pvalue 0 ex could difference between 2 means be as large as it is b chance 0 Could of outcomes be as large as it is by chance alone EX we have hypothesis that someone has ESP can predict the future Null hypothesis the person has no special abilities and CANNOT predict future Research hypothesis the person DOES have ESP and can predict the future Data collection we have the person predict the outcome of one coin flip Say they get it right does this prove they have ESP What is p vaue for our result probability of getting this result if the null hypothesis is true p50 Suppose the person predicted 10 consecutive coin flips correctly What is the p va For this result P00O98 Comparing two means Null hypothesis population means are any differences between sample means are due to CHANCE random error Research hypothesis H1 pop Means are not T test test statistic associated with a probability of obtaining sample means that differ by observed amount if population means were The t test T difference between groups means normal variability win groups Ift is large the difference between groups is BIGGER than normal variability win groups 0 Therefore 2 groups are significantly different Ift is small the diff between groups is SMALLER than the normal variability within groups o Therefore 2 groups are NOT significantly diff Analysis of variance ANOVA computes the p vas for main effects and interactions in factorial designs F test is an extension of the t test o Computes ratio of systemic variance deviation of group means form overall mean to error variance deviation of individual scores in each group from group mean 0 Similar to t test that it compares differences between groups to variability win groups F test asses probability that main and interaction effects could be due to chance alone Type 1 and type 2 errors When assessing statistical significance 2 types of errors can make based on yesno significant decisions Type 1 error incorrectly rejecting the null when it is CORRECT O O Gullibility If concluded someone was psychic when they are not based on one coin flip committed type 1 error incorrectly rejected null Here p value was 5 The lower the p va considered significant the less likel to reject the null hypothesis and less likely to commit type 1 error Type 2 error incorrectly accepting the null hypothesis when it is in fact false blindness to a relationship 0 0 Say someone IS psychic and correctly predicts 3 straight coin flips pva P125 may not be improbable enough to say statistically significant don39t reject the null How could we avoid the type 2 error in this case while finding an acceptable pvalue As sample size increases the probability of a type 2 error decreases more likely to detect an effect at higher sample sizes larger sample sizes have more power to detect effects that are present low power in the 3 coin flip example Alpha the p vaue at which we decide to reject the null O as alpha qets larger the probability to a type 1 error increases and the probability of a type 2 error decreases alpha 05 is often used in psychology research as compromise between the odds of type 1 and type 2 errors p vaues are not the same as effect sizes 0 correlation of r80 indicates a strong relationship between 2 variables but r10 is a weak relationship a p va for correlations indicates the probability of finding a correlation as large as that observed if the true correlation is 0 with a large sample size may have a small p val for r10 simple confident that this greater than 0 but this is still a a weak relationship very low p va p00001 simple express confidence that findings are not due to chance but say nothing about the size or practical importance of the finding hierarchy of internal validity 1 true experiments post test only pretestposttest repeated measures matched pairs 2 quasi experiments one group pretest posttest nonequivalent control group nonequivalent control group pretest posttest 3 correlational designs ex say we want to test whether PowerPoint or chalkboard lectures produce higher scores Instructor uses PowerPoint in psych 7 in one quarter but chalkboard in the next quarter and compare exam scores What type of design say chalkboard scores were higher Alternative explanations external validity 311 the extent to which we can generalize our results measures the extent to which research findings can be generalized to other participants settings variables big issue for psych research most research is conducted with college students in western cultures even if a study has strong internal validity would get the same results with other participants 0 Interaction between the factor being investigated and the type of participant Use factorial designs to test generalization Culture of honor study ypothesis men from the south would respond differently to being insulted experimental manipulation than men from the north theory south have culture of honor men are expected to retaliate strongly against threats or insults or they lose status w fam friends economy in south historically based on herding culture of honor passed down through some cultural transmission process possible explanation for higher rates of retaliatory homicide in the south In response to insults hypothesis interaction effect of regional origin and insult with southerners reacting with more aggression after the insult than northerners study 1 design 2x2 design w participants region of origin n vs s as one variable and condition insulted vs not as other variable participants 83 Michigan undergrads 42 from north 43 from south lived in south for at least 6 years all white all non Michigan residents dependent variables multiple measures word completion exercise rate how angry ppl look how many ppl end stories with violence study 1 results word completion observer ratings ambulance story face ratings NO interaction how would we test a main effect of growing up in north vs south Find the averages of south and north Was there a significant difference for insult vs control for southerners only 0 Simple main effect They were not effected Study 2 Same procedure but with more dependent measures Dependent variables 0 Cortisol responses stress hormone saliva samples collected before and after manipulation o Testosterone responses pre and posttest saliva samples Results from 1 and 2quot studies Southerners more affected by insult than northerners for measures of 0 Using violence to end a story 0 Cortisol increases 0 Testosterone increases Results provide evidence for cultural differences in reactions to insults Alternative explanations o Other than culture of honor 0 Better manners more insulting to them 0 Depends on the setting Replication and external validity Specific studies are performed in specific places an times with particular operational definitions of variables Replication of results in different settings w diff participants or diff operational definitions increase generalizability external validity of results 0 Exact replications 0 Conceptual replications Different operational definitions can est boundary conditions of theoretical constructs Distinct results w diff procedures suggest alterations to theory Conceptual vs exact replication study looks at effects of extrinsic rewards on math performance study uses specific rewards specific types of math problems in a specific sample of students operational definitions of each variable an EXACT replication would employ the same operational def s but use new subjects 0 test how consistent the findings are a conceptual replication will change the aspects of the design while testing the same general idea 0 may change the size of rewards types of problems length of retention o changing operational definitions to test how generalizable effect of extrinsic rewards are review articles as assessments of external validity qualitative reviews meta anaysis gender and aggression o quantitative compute average effect size across many studies 0 can compute different effects under different circumstances verbal vs physical aggression file drawer problem hard to publish results where there Is no significant effects my experience with external validity testing results with different subject populations study design male university of Chicago student randomly assigned to condition 5 minute convo saliva sample take before and 15 minutes after interaction randomized pretest posttest design subjects rated confederates for physical attractiveness 9 external validity problem UCSB students are very different from Chicago Ucsb study Male ucsb students randomly assigned to male experimenter or female Saliva samples before and 40 minutes after Results men who talk to women have sig increase FINAL REVIEW 50 mc question 1115 115 Tuesday alex office hours theory 9 hypothesis 9 predictions 9 data 9 conclusions when doing a study reliability will this be consistent o Test retest stability overtime give test on day 1 and then two weeks later are they correlated o Interreliability coding system two ppl can agree on Internal consistency reliability 0 How different are the sub components Intelligence tests subtests g o Splithalf reliability 0 Chronbach Alpha inner correlations between items on test High correlation measuring the same thing good Validity is it doing what we think it is doing 0 Faceworthy does it seem like it is doing what we think it should Measure of male genetic quality fluctuating asymmetry Doesn39t have to be face valid to be valid 0 Predictive validity Do The scores predict behaviors that should be correlated or related to 0 Criterion group validity o Concurrent validity do specific populations score on the measures as they would be predicted to score Psych professors expected to do worse than chem Sky divers and librarians o Divergent o Convergent o Discriminant when the measure is unrelated to the thing you wouldn39t expect it to Correlation Democrats and environmental concern Positive linear relationship R9 0 Straight line no relationship Clarifying vs alternative explanations Ice cream 9 cramps 9 drowning Ice cream causes cramps which causes drowning Temperature is a clarifying variable Basic experimental designs Posttest only only measure the dependent variable once Pretest posttest design measure dependent variable then manipulate then test dependent variable again Repeated measures design each subject goes through every condition even factorial designs Matched pairs design book or computer program example Do intelligence test and match everyone based on IQ s and scores in the class and split them apart Dealing with order effects Complete counterbalancing present the different conditions in every possible order For 2 conditions controlexperimental only 2 possible orders 4 conditions 24 distractor taskstime lags present one type of stimulus but then have rest period avoid fatigue latin square designs rule 1 each condition appears at each ordinal condition rule 2 each condition precedes and follows each condition one and only one time factors and levels factors variables which can contain numerous levels caffeine sex levels within each factor caffeine level 19 0 mg level 2 9 50 mg level 3 9 100 mg sex level 1 9 men 0 level 2 9 women main effect looking at just one factor how many effects can there be for the factor caffeine with 3 levels 1 what kind of relationship could you test for with this design that you couldn39t with only 2 Curvilinear nonmonotomic relationship like a U Alcohol study in factorial notation Main effect no main effect American idol slide main effect yes because 150 is different from 250 simple main effect looking within the square graphing results interaction effects plot all data as line graphs nonparallel lines indicates interaction comparing two means null hypothesis population means are Any differences between sample means is due to chance random error Random error everything that is not in our manipulation research hypothesis alternative population means are NOT t test test statistic associated with probability of obtaining sample means tat differ by observed amt if population means were null hypothesis is true 0 use p vaue pvalue statistical significance assesses the probability that results could be due to chance rather than hypothesized cause could difference between 2 means be as large as it is by chance Where chance is everything not accounted for in manipulation How likely we get a true or real difference between groups
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'