656 Chapter 17 Log-Linear Analysis
Motor
Vehicle Helmet Injury Count
No
Ye s 2 0
No 195
Old Ye s
Ye s
Ye s 2
No 17
No
Ye s 4 2
No 174
No
Ye s
Ye s 2
No 43
No
Ye s 3 0
No 421
17.5 Examine the table of cell frequencies and predict what terms must be in the model to
explain the pattern of differences.
17.6 Run the loglinear analysis using the saturated model and draw conclusions
17.7 Now run the analyses using SPSS HILOGLINEAR to solve for an optimal model using
backward deletion.
17.8 What is the optimal model calculated in Exercise 17.7?
17.9 Interpret the results of your analyses.
17.10 Maimaris et al. concluded “No child who wore a helmet at the time of the accident sustained
a head injury.” Does that seem like a conclusion that would follow from your analysis?
17.11 Calculate the odds ratio of an adult having a head injury as a function of whether or not that
adult wears a helmet. (Why could you not do the same thing for children?)
Dabbs and Morris (1990) investigated the effects of elevated testosterone levels in a repre-
sentative sample of U.S. adult males (mean age 37). Subjects were classified as High (upper
10%) or Normal on testosterone, as high or low on socioeconomic status (SES), and as en-
gaging (or not engaging) in adult delinquency. Their data follow.Low SES High SES
Normal High Normal High
Delinquent Testosterone Testosterone Testosterone Testosterone
Yes 190 62 53 3
No 1104 140 1114 7017.12Calculate the odds of being classed as an adult delinquent for each of the categories in the
preceding table.
17.13 What are the odds ratios of delinquency for the four SES/Testosterone groups in the preced-
ing table?
17.14 Apply a log-linear model to the data from Dabbs and Morris and interpret the results.
17.15 Dabbs and Morris collected data on a number of other variables, including childhood delin-
quency, hard-drug use, and many sex partners. Why would it be inappropriate to create a
dimension labeled Behavior (adult delinquency, childhood delinquency, hard-drug use,
many sex partners) and use that as an additional variable in the analysis? In other words,
what is wrong with analyzing SES 3 Testosterone 3 Behavior?