Homework Assignment #5

All homework assignments must typed or legibly handwritten, and must show your name, the assignment number, course number, and the date clearly.  Show all your calculations and other work.  Homework is due at the beginning of class on the specified date.

Due Monday, March 19

Chapter exercises:

ch. 6: 8 (also compute RR and OR for part b and interpret), 14
ch. 7: 12

Additional exercise:

1. On the first day of class, you and your classmates completed an anonymous survey.  Students were randomly assigned to complete one of two versions of the survey.  The two versions differed only with respect to the response options for the question about how many hours you spent per week studying for a typical course.  One version had response options with narrow intervals biased toward the low end (1-2, 3-5, 6-8, 9+) and the other had response options with wide intervals biased toward the high end (1-8, 9-12, 13-16, 17+).  The goal for the experiment was to see whether the different versions (with the different response option sets) influenced the responses you gave.

The following crosstabulation describes the results:

                                  # hrs spent studying per week

response options	low [1-8]	high [9+]	total
narrow	11	0	11
wide	7	4	11
total	18	4	22

(Interestingly, the 4 students who reported a high level of studying selected the "13-16 hours" option).

Compute the relevant percentages within each category of the independent variable.  Calculate Goodman and Kruskal's tau and interpret the results.  Why can't you compute the relative risk or odds ratio in this particular case?  What can you conclude about the influence of response options?

SDA task:

Pick two ordinal variables that you're interested in (from any data set).  Specify the survey, and the meaning of the variables.  Define the independent and dependent variables for your analysis.  Using the tools in SDA, compute the relevant row or column percentages (depending on which you think should be the independent variable) and compute the appropriate measure of association (check the "Statistics" box in the Frequencies/Crosstabulations screen).  Be sure to offer an interpretation of the results.