Exam1 PS 306, Spring 2006 1. Dr. I. P. Freeley was interested in replicating the Middlemist et al. study (effect of invasion of personal space on time to urinate), but using a repeated measures design for greater power. For several days, he had a rotating cadre of confederates (so that the students wouldn’t think that someone was stalking them) follow a set of students enrolled in an introductory psychology class. Whenever one of these students would enter a restroom to urinate, a confederate would check to ensure that no one else was using a urinal. If the participant were alone at one of the urinals, the confederate would either: 1) go to the urinal immediately next to the student (Near Stall); 2) go to a urinal one urinal away from the student (Distant Stall); or would simply go to the mirror and comb his hair (Alone). The dependent variable, as in the Middlemist study, was the time (in minutes) between when the unwitting participant unzipped his pants and when he began to urinate (micturate). Complete the source table below and interpret these data as completely as you can. [10 pts] 9 1.228 .1362 2.282 1.141 111.217 <.0001 222.433 1.00018 .185 .010DF Sum of Squares Mean Square F-Value P-Value Lambda PowerSubjectCategory for DistanceCategory for Distance * SubjectANOVA Table for Distance 10 .550 .227 .07210 .560 .196 .06210 1.140 .259 .082Count Mean Std. Dev. Std. Err.AloneDistant StallNear StallMeans Table for DistanceEffect: Category for Distance There is a significant effect of the Presence/Distance of another person, F(2,18) = 111.217, MSE = .01, p < .001. Post hoc analysis was conducted using Tukey’s HSD: ! HSD = 3.61.0110= .11 Men with another man at the near stall took significantly longer to begin urination (M = 1.14) than men with another man at a distant stall (M = .56) or men who were alone in the restroom (M = .55). 2. Mook argues that external validity is not always the purpose behind psychological research. For each of the studies below, indicate why the study is not externally valid, then why it’s not a concern, given the intentions of the researcher(s). [10 pts] Use Mook article to respond to this question. Study Why not externally valid Why lack of EV is not a concern Argyle (glasses and intelligence) Harlow (infant monkeys and drive reduction theory)Hecht (dark adaptation) Brown & Hanlon (parental role in grammar acquisition) 3. Correlational designs do not allow you to make casual claims. Why not? Be very explicit about the difficulty of claiming that changes in one of the two variables in a correlational study causes the related changes observed in the second variable. We also discussed the shortcomings of using non-manipulated characteristics of the participants as “independent variables” in an experiment. How is this class of variable related to the notion of correlational designs? [10 pts] In class, we discussed the three possibilities for an observed significant correlation between two variables (X and Y). One possibility is that X does cause Y. Another possibility is that Y causes X. And, finally, there’s the possibility of a third variable (Z) that might cause both X and Y. And of course, we’re not talking about a host of potential third variables (Z1, Z2, Z3…). Each person carries around a whole host of characteristics (resulting from nature and nurture). Thus, when a researcher looks at such characteristics as an “independent variable,” he or she runs the risk that the characteristic is not actually having a causal impact on the dependent variable. Instead, some “third variable” that is related to the nominal characteristic may be actually causing the behavior observed in the dependent variable. For example, suppose that a researcher were interested in studying the impact of IQ on performance in a public speaking task. It may well be the case that low IQ causes people to perform more poorly on such a task and high IQ causes people to perform better on the task. However, it may be that other variables related to IQ may be the causal agents. For instance, people with low IQ may have lower self-esteem, which may lead them to be more uncomfortable when speaking in public. Or it may be that people with high IQ are asked to give their opinions more often, so they have more experience speaking in public. In essence, then, studies using non-manipulated characteristics of participants as “independent variables” are not true experiments, but actually correlational, because the “independent variable” is not being manipulated. And then, of course, there’s the definite possibility of a third variable potentially causing any observed effects. 4. Two researchers were interested in studying the effects of reward magnitude on performance. Both researchers used introductory psychology students as participants, the same total number of participants (21), the same type of reward and reward magnitudes ($1, $5, $20), the same apparatus, the same task, and the same performance measure (DV). One researcher used an independent groups design and, on the basis of the results, cannot reject the null hypothesis (that reward has no effect on performance). The other researcher used a repeated measures design and found a statistically significant effect of reward magnitude—larger rewards lead to better performance. Assume that neither study has a major flaw (e.g., repeated measures design is properly counterbalanced, random assignment to conditions). There are two fundamental reasons why the two researchers might have reached different conclusions. One reason concerns the sensitivity of the test of the null hypothesis. The other reason concerns the nature of the participant’s experience in the two studies. Provide me with a clear explanation of the two reasons for the different results that the two researchers obtained. Would you trust the results of one study more than the other? Why? Finally, complete the source tables for the two experimenters seen below. [10 pts]Independent Groups Design (FCrit = 3.55): Source df SS MS F Treatment 2 28 14 3.5 Error 18 72 4 Total 20 100 Repeated Measures Design (FCrit = 3.23): Source df SS MS F Subject 20 100 5 Treatment 2 20 10 5 Error (Subj x Treat) 40 80 2 Total 62 200 5. In your first lab, there were a number of different personality measures. One was the Rosenberg Self-Esteem Scale (selfestm) and another was the Satisfaction with Life Scale (swls). Had you