3100 Test Two Study Guide
3100 Test Two Study Guide 3100
Popular in Advanced Experimental Psychology
verified elite notetaker
Popular in Psychlogy
This 10 page Study Guide was uploaded by Grace Gibson on Thursday March 3, 2016. The Study Guide belongs to 3100 at Clemson University taught by Dr. Thomas Britt in Winter 2016. Since its upload, it has received 66 views. For similar materials see Advanced Experimental Psychology in Psychlogy at Clemson University.
Reviews for 3100 Test Two Study Guide
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: 03/03/16
3100 Test Two Assumptions Participants Make in Research ● Maxim of Relations ○ participants want to make sure they are contributing to the research ○ participants take into account the context of the questions ● Maxim of Quantity: participants will give information researchers are interested in, and no more ● Maxim of Manner ○ participants will assume the researcher has a purpose in designing the questions ○ contributions will be clear ● Demand Characteristics: they say what you want to hear Ethical Concerns in Research ● Milgram’s obedience study ● Zimbardo’s prison study ● Micturation Delay study ● Humphrey’s study of gay men ● Research using deception is prevalent in research ○ there’s a lot of standpoints on whether or not deception is okay ○ deontological position: any deception is unethical ● an interesting finding is that people enjoyed the study more when they were deceived ● utilitarianism: comparing the benefits of the research with the cost ● cost/benefit analysis is the position adopted by the APA ○ a joint commitment to protecting the rights of the participants while advancing knowledge ● Debriefing: after the study, educate the participants on the study ○ tell them about any deception and why it occurred ○ ask them about their response to the study ○ address any concerns the participants have ● Process Debriefing: more exact (ask them about their logical process) ○ often necessary when there is major deception or when the participant engages in unexpected behavior ○ necessary for the receipt of performance or interpersonal feedback ○ may need to be an interactive process rather than a survey Conducting Your Research ● treat participants with respect ● be professional at all points in the study ● make sure directions are clear and easy to follow ● maintain consistency across all participants ● make sure you are completely “blind” to the experimental condition ● following a script helps you accomplish all of these ● IRB ○ identify the purpose of your research, potential benefits and costs to the participants, and how you will address these costs ○ how will participants be recruited? (no coercion is allowed) ○ participants have to complete an informed consent form Correlational Research Correlational Research ● correlational research defines features of interest ● how is change in one variable associated with another variable? ● measured variables can be continuous (e.g. intelligence) ● measured variables can be discrete (e.g. gender, ethnicity, religion, etc…) ● correlational coefficients do not have to be the statistic you use ○ both variables are continuous: correlational coefficient ○ one variable is continuous, one is discrete: ttest of ANOVA ○ both variables are discrete: chisquared ● one variable is assumed to be the cause of the variation in the other ○ predictor: causal variable (IV) ○ criteria: outcome variable (DV) ● Causal Inference Problem: a correlation does not mean one causes the other ● Third Variable Problem: concern that another variable might account for the statistical association between two variables of interest ○ Americans noticed that cases of polio increased when people were eating more ice cream and soda ○ polio cases were actually increasing because it was hot (and therefore people were eating more ice cream and soda) ● correlational research is the most common form of research in psychology ● try to follow up with experimental research when possible Issues in Correlational Research ● correlational research only examines linear relationships so if you’re looking for a nonlinear one, you’ve got to do something else ● your theory may posit another idea that is not linear ● there are formulas you can use to test for other relationships ● e.g. Arousal/Performance Relationship: a median amount of stress produces the best performance ● Leary was interested in selfesteem as a function of interpersonal feedback ○ he predicted an “ogive” relationship ○ (example just to see the shape of this relationship) ○ if he had done a linear test, he wouldn’t have found this relationship Interpreting Correlational Findings ● r = 0.10 r^2 = 0.01 ○ small correlation ● r = 0.30 r^2 = 0.09 ○ medium correlation ● r = 0.50 r^2 = 0.25 ○ large correlation ● r^2 is the percentage of the variation in one variable that is accounted for in another variable (proportion of shared variance) ● when interpreting the size of the correlation, consider: ○ the size predicted by a given theory ○ whether variables are assessed with different methods ○ size of the sample (p value is based on sample size) ○ even small correlations can be meaningful (aspirin and heart attacks) ● Restriction of Range: due to the sample, one or both of the variables fail to show variation ○ similar to ceiling and floor effects ○ e.g. SAT scores vs. freshmen GPA ■ people who get into clemson have pretty high SAT scores ■ freshmen will have better GPAs because classes are easier ■ so there won’t be much variation because of the sample Mediation and Moderation ● mediator: transmits causality ○ indicates the cause of the relationship ○ so two things are correlated, why are they correlated? ○ e.g. research has shown that people who are highly conscientious have higher job performance (mediator: they set specific goals) ○ e.g. research has shown that people of higher economic class do better in school than people of lower economic class (mediator: presence of academic role models) ○ if you control the mediator and the correlation stays the same, the mediator has no effect ● mediation: examining what accounts for the relationship between two variables ● moderation: the relationship between two variables depends on the level of a third variable ● moderator: changes the relationship between variables ○ moderation indicates the strength of the relationship ○ e.g. what buffers individuals from the negative effects of stress? (moderator: coping abilities) ○ e.g. research has shown a relationship between combat exposure and PTSD (moderator: morale) Advanced Correlational Methods ● longitudinal studies (prospective designs): measure the same variables at multiple points in time and look at the pattern of correlations ● multiple regression: use of more than one variable to predict another variable ○ predictor variables vs. criterion variables ○ e.g. SAT scores vs. freshmen GPA ○ you get better prediction with predictors that account for unique variance ○ R = correlation between set of predictors and criterion ○ R^2 = percentage of variance accounted for in outcome by predictors ○ a lot of people say you can’t say a variable predicts another variable unless there is a high correlation ○ standard multiple regression: all variables go in at one time to predict variance ■ do all variables predict the independent variability of the outcome? ■ this probably won’t happen if the two variables overlap (e.g. SAT scores and intelligence when predicting GPA) ■ usually for small numbers of variables (four or less) ○ stepwise multiple regression: variable with the highest correlation goes in first followed by all the other variables that account for unique variation ■ lots of variables ■ tackles one variable at a time ○ hierarchical multiple regression: order of the variables is based on theoretical reasons ■ usually used if you have a set of controlled variables you want to put in first and then you put in a set of your theoretically related variables ○ logistic multiple regression: used if you have a categorical (binary) outcome ■ e.g. received treatment or not, graduated or not, etc… ■ focuses on accuracy of predicting two groups ● Factor analysis: how many dimensions underlie a large number of variables? ○ helpful in a theory that posits an underlying structure of a construct ○ maybe you’re doing a study on stress and you see that the factors are underlied by four dimensions (academic, physical, social, and emotional stress) ○ factor analysis was used to determine the Big Five Personality Traits ○ used to determine factor underlying our evaluation of things (we evaluate things as good/bad, fast/slow, weak/strong) ○ uses a scree plot to determine the number of factors ■ then you rotate the factors to see which items load on factors (this number is basically the correlation between item and the factor) ■ simple structure: item loads 0.40 or higher on only one factor ■ if there are items that load 0.40 or higher on more than one factor or that don’t load on any factors, get rid of them ■ Eigenvalue: Experimental Research Design Experimental Research ● you manipulate your IV and assign random people to random conditions ● control/baseline groups are important ● control group: gets exactly the same treatment as the experimental group, but no touch of the IV ● McCord studied the effect of positive or negative feedback on people’s selfesteem and found that the control group had better selfesteem Placebo Effect ● only 10% of people on antidepressants are getting the true drug benefits ● if you want to see if a drug isn’t just effective because of a placebo effect, you would have to do a blind study (experimental group, control group, and group of people who do no drugs) ● maybe include a group that’s told their taking the fake drug (sugar pills) but they’re actually taking the Prozac ● this shows the importance of a control group ○ it refers to the tendency for people to report a treatment has benefitted them, regardless of whether the therapeutic agent is actually present) ● spontaneous remission: participants go through no treatment and are aware that they are not receiving any treatment but still get better ● there was an experiment that looked at drug treatment and how it affected public speaking ○ one group didn’t do anything, one group had a placebo, and one group actually got the treatment ○ 22% had spontaneous remission, 28% had the placebo effect, and 35% benefitted from the therapy ● Hawthorne Effect: researchers investigated the impact of higher or lower lighting on job performance ○ both raising and lowering the lights improved performance ○ concluded that if people believe they are being intervened with, performance will improve ○ this also proves that the control group is important Manipulation Checks: Testing Validity of IV ● just because you think you are manipulating your IV, doesn’t mean you are ● you need to insure that your manipulation was effective ● manipulation check: questions completed at the end of the experiment to assess the effectiveness of the manipulation of IV ● e.g. if you gave positive feedback to one group and negative feedback to another group, ask they at the end of the study how they feel they did on the test ● we want significant and large differences between conditions on the manipulation check ● Britt did an experiment where they told some people they might get a shock to the eye to study anxiety ○ there appeared to be no difference in anxiety levels between the group who thought they might be shocked and the group who didn’t think they’d be shocked ○ they did not actually manipulate the IV ○ if they hadn’t done this manipulation check, they would have falsely made a conclusion ● if you haven’t really manipulated the IV, you aren’t really measuring its effect on the DV ● you can do a pilot study where you insure that your IV can be manipulated Designing the Experiment ● you want to manipulate and control one key variable ● it is critical that only your variable of interest is manipulated ● this does not mean you can’t have multiple IVs ● error variance cannot be avoided completely since there are many factors that will affect your DV that have nothing to do with your manipulation of your IV (personality, mood, time of day, etc…) ● it is important to understand that the IV accounts for part of the variability of the DV and this is fine as long as the error variance is spread equally among conditions ○ random assignment ensures this ● confound: a factor that is unintentionally manipulated along with your IV that may be responsible for the relationship between the DV and IV ○ e.g. he assigns the first 3100 lab to get time management training and the second 3100 lab as the control group ○ he compares their end of semester GPA in the course ○ he has manipulated the lab time by not gathering each group from each lab ○ this was not random assignment Internal Validity (basically the opposite of a confound) ● when the internal validity is high, we are confident the differences in the DV are due solely to our manipulated IV ● threats to internal validity include biased assignment, differential treatment, differential attrition, historical influences, and demand characteristics ● biased assignment of subjects means there isn’t random assignment ○ there is a possibility the groups differ on attributes other than the IV ○ Langer and Rodin study: they assigned one floor of the nursing home to have active control of their schedule and another floor was the control group ○ they were trying to see if patients would live longer if they had more control over their schedule ○ they didn’t account for the differences that could occur between floors ● differential treatment of subjects in different conditions is the most frequent problem if there is true random assignment ○ subjects are treated different in addition to the IV ○ e.g. if you’re delivering positive or negative feedback or looking at self esteem, you might try to cushion the blow of the negative feedback or act more excited when delivering positive feedback ● differential attrition in conditions usually happens if you have a long and strenuous study ○ if you have different numbers of people who leave your study, the ultimate concern is that you have a personality difference between groups ○ e.g. if one group is more demanding than the other, more people might leave this group and the people left might be more hardy or something ● historical influences on the IV ○ there could be a historical event that influences the effect of the IV on the DV ○ e.g. Magic Johnson coming out about his AIDs and how that affects public opinion ● Demand characteristics may be magnified in withinsubject designs ○ you need to make sure your participants aren’t figuring out what your hypothesis is ○ they may try to help or hurt your hypothesis ○ you can ask questions at the end of the study to avoid this Questions when Designing an Experiment ● Are you interested in the effects of one variable or more? ○ one IV is a one way design ○ more than one IV is a factorial design ○ levels in the first IV times levels in the second IV equals the number of conditions ○ you can draw out your design (3x2, 4x3, etc…) ○ how many levels your IV has depends on your hypothesis ● Do you want different people in different conditions or the same people in all conditions? ○ between groups: different people ○ within groups: same people (also called repeated measures ○ split plot design: one or more variables between and one or more variables within ■ e.g. soldiers go through either stress or mindfulness training and are questioned before and after ○ Advantages of within group ■ this removes the error that results from individual differences in personality ■ you need less people ■ because of these advantages, if you can do within group you should ○ Disadvantages of repeated measures ■ carryover effect: exposure to one level of the IV can affect their response to another level ■ fatigue might set in ■ participants might figure out the purpose of the experiment ■ people will eventually habituate to the DV ○ counterbalancing is a way to partially control for carryover effects (e.g. you can vary the order of levels of the IV for different people) ● Do you want to measure the DV before you manipulate the IV? ○ this is called pretest/posttest design and it is necessary when you are examining change ○ however, the pretest may sensitize participants to the purpose of the study ○ the pretest should not immediately precede the manipulation of the IV ● Do you want to be sure the subjects are equal on variables related to the DV? ○ this is called matched random assignment: divide the subjects on an attribute, then randomly assign them to conditions ○ this is often not done because researchers just trust random assignment ● Do you want to measure one variable and manipulate another? ○ this is a mixed model design (expericorr) ○ this is useful if you’re interested in how individual differences interest with the manipulation to influence the DV ○ how to create groups from a continuous variable: you can use the median to divide into low and high groups (median split) ■ however that means people around the middle will be in different groups even though they are similar ■ you can take the top 25% and the bottom 25% and make extreme groups (however, this means you lose 50% of your people) ● Do you want to test for the significance of the IV after controlling for another variable? ○ Analysis of Covariance ○ this can be used to test why an IV is causing variation in a DV ○ e.g. do nature sounds or silence improve math performance? ■ you may think nature sounds relax people so they do better ■ you can measure relaxation and control that ■ this tells you that if you removed this variable, the IV would still have the same effect on the DV ■ you could include relaxation as a covariate ○ if your IV is not longer significant, you have support for mediation ○ someone might do an analysis of covariance to decrease error in the study ○ you want your covariance to be related to your DV but not your manipulation External Validity of an Experiment ● Will the relationship between the IV and DV be found in the real world? ● some people say we shouldn’t worry about this and just control variables the best they can and really study the IV and DV ○ this is done in biology and physiology ○ you have to have a strong IV ● it is possible to test effect size in a laboratory vs. field ○ there was an analysis conducted with findings from social psychology ○ the correlation between effect size in the field and the laboratory is 0.75 ○ a bigger effect in the lab leads to a bigger effect in the field (at least for social psychology) ● don’t feel guilty if you feel like your study is artificial Programmatic Research ● you rarely have important findings through one experiment ● programmatic research: conducting multiple studies involving different styles and techniques ● this can help you identify underlying processes responsible for the effects ● optimistic bias: tendency to believe we are less likely than the average person to experience medical illnesses ○ we also bias the interpretation of feedback ○ it has been found that part of the prefrontal cortex is responsible for this bias ● programmatic research is very time consuming QuasiExperimental Design ● the IV is naturally experimented (not by you) ● the experimenter doesn’t have control of the IV ● there’s nonrandom assignment ● this creates the possibility for confounds ● there are three types: nonequivalent control group design, natural experiments, and time series design ● in a nonequivalent control group design, the manipulation is a program of some kind ○ the researcher has control over the active manipulated group, but not the control group ○ the nonrandom assignment is basically what makes this a quasiexperimental design ○ e.g. a school psychologist that is testing the effect of a drug intervention program on drug use might use their school as the manipulated one that goes through the program and another school as the control school ○ the control group might be the same on some key demographics, but you can never really be sure this control group is the same on all dimensions ■ is possible these differences between groups can be confounds ○ pre/post design: looking at drug use before and after the intervention (just looking at the manipulated group; there’s no control group) ○ one time study: measuring both groups after the program ■ you’re not taking into account what the schools were like before the program ○ if you have both groups and do pre/post test for each, you’re controlling for all these factors ● natural experiments are the results of events like terrorist attacks, hurricanes, etc… ● time series design
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'