This exam is based on a single study by Iyengar & Lepper (2000). You may not look at that paper before you turn in your exam, but you are welcome to look at it later. [The first author was originally Sheena Sethi, whose work you will come across in the book in connection with a study on religious fundamentalism. (Well, "fundamentalism" has taken on a new meaning these days.)]

Iyengar and Lepper were interested in studying what happens when we present people with choices. They suspected that some choice is a good thing ("Would you prefer this Hershey bar or those M&Ms?") but that too much choice can be just that--too much (Which of these 32 different candy bars would you prefer?)

By way of background, the study discussed here was Study II of a set of three studies of this phenomenon, (The first study compared choosing among 6 jars of jam versus choosing among 24 jars. The third study involved choice of Godiva chocolates.) In this study (Study II) all 193 students in a Social Psychology course were required to watch "Twelve Angry Men," and then were given the option of an extra-credit assignment of writing a response-paper to the movie. The students were given a list of paper topics, and this list either included a choice of six different topics (Limited choice) or a choice of 30 different topics (Extensive choice). The experimenters recorded how many students in each condition chose to write a paper, and a measure of the quality of that paper. Quality was judged in terms of both Content and Form on a 10 point scale. (Interestingly, the students were told in advance that they would get the extra credit just for handing in a paper, and that the quality of the paper would not matter. However, it is still of interest to see which students turn in better papers, and whether the quality in some way relates to the number of choices they were given.)

You can either pick up the data on a disk from Diana or Mary (copy and return the disk for the next person), or you can open the web page for this exam (It will no longer be last year’s exam.) and copy if from there, or you can open Gumby (if you are in Waterman lab) and get it from there. The file is named Midterm01.dat.

I am presenting the data in the form of an ASCII file. This is a simple text file with one line for each subject, and spaces between the individual variables. I have chosen not to put the variable names in Row 1, and leave that task to you after the data have been read by SPSS. The variables, in order, are ID, the student’s Gender, the Choice condition the subject was in (Limited vs. Extensiv), the student’s Grade on a previous midterm, whether the student Wrote a paper (Yes vs. No), the rating of the Content of the paper, and the rating of the Form of the paper. They also combined the Content and Form ratings into an overall Total Quality score (TotQual). You will not need all of these variables, but I have included them to make the data have the kind of structure that we often see when analyzing data from a real experiment. Although the questions below are specific to certain variables, I give you the opportunity to work with additional variables.

When you look at the data, you will discover that I have used string variables in several places, whereas in lab we used numeric variables almost exclusively. When you read the data into an SPSS file, you will have to take this into account. Be sure that you are careful with spelling and capitalization. You may have some difficulty reading these variables in, but if you work carefully and read the choices in the dialog boxes, you should be OK. You may want to clean up the data sheet when you get it, so that things look less confusing, and you will certainly want to name your variables. (If you go to the "variable view," you will be able to set the number of columns (not "width") in the display and the alignment. (You may have to scroll to the right to find these settings.) You will see that there are missing data. If a person did not do the extra credit assignment, the authors could hardly grade it for form and content. The missing data should show up as a "period."

Each question, or subquestion, below is worth 5 points, except for 6a) and 6b), which are each worth 10. (I know that adds to 110 points.)

1. Describe the data using appropriate descriptive statistics and graphics. Are you satisfied with what you see?

2. I tried to make the data look skewed, but was not successful. What would you have expected that the distributions would look like in terms of the shape of the distribution?

3. If I had been successful in creating skewed data, and if they were very positively skewed, what might you do to make them somewhat more symmetric?

4. The authors hoped to show that people were more likely to write the extra-credit paper if they were faced with fewer, rather than more, choices.

a. What would be the null hypothesis corresponding to their prediction?

b. Set up a results table to show what they found.

c.

d. What statistical test would you use in this case, and why?

e. Apply that statistical test, using SPSS, and paste the output here.

f. What would you conclude from the answer to part d?

g. Write a short paragraph, as you might expect to find in their results section, telling me what you found and what you conclude.

5. The authors also suspected that for those students who chose to turn in a paper, the number of choices would play a role in the quality of papers the students turned in (if they turned in any).

a. What hypothesis or hypotheses would you expect as their research hypotheses?

b. What would be the null hypothesis(ses)?

c. What would be an appropriate statistical test for the quality of the papers as a function of treatment condition, and why?

d. Use SPSS to run the test (or tests) and paste the output here.

e. In order to make sense of what SPSS has printed, you need to attend to group variances. Tell me what that statement means, in terms of what you would look at, and tell me what conclusions you would draw.

f. Your printout will have two different tests for each dependent variable. Why do those tests have different degrees of freedom. Which result is appropriate for this particular set of data?

g. I believe strongly that it is most appropriate to run a two-tailed test here.

a. Why should the test be two-tailed?

b. What would you do if I said that I wanted you to run one-tailed tests?

6. So far we have been looking at this problem from the point of view of null hypothesis testing. If you sent a paper in this way, I would hope that the editor would return it to you with the request that you say something about the size of your effect. We have talked about a number of statistics that you could use, and have seen examples in lab.

a. With respect to whether or not someone chose to do the extra-credit assignment, what statistic(s) might we use to satisfy the editor? If you have more than one, do they tell you the same kinds of things, and are you equally happy with them?

b. With respect to the measures of the quality of the papers that were turned in, what measures might you use, and what are their values? (Hint: The text talks about "effect size," "squared point-biserial correlation," and "confidence limits." So should you.

7. What role do random sampling and random assignment play in this study?

8. For up to 5 points worth of extra credit, think about what other questions you might ask of these data, and tell me what you find. (Remember, it is only worth 5 points, so doing every analysis you can dream up doesn’t seem like a good idea. Be selective.)