Statistical Methods for Psychology

314 Chapter 10 Alternative Correlational Techniques

10.17 On page 302 I noted that Rosenthal and Rubin showed that an of .1024 actually repre-

sented a pretty impressive effect. They demonstrated that this would correspond to a of

20.48, and with 100 subjects in each of two groups, the 2 3 2 contingency table would

have a 34:66 split for one row and a 66:34 split for the other row.

a. Verify this calculation with your own 2 3 2 table.

b. What would that 2 3 2 table look like if there were 100 subjects in each group, but if

the were .0512? (This may require some trial and error in generating 2 3 2 tables

and computing on each.)

10.18 Using Mireault’s data on this book’s Web site (Mireault.dat), calculate the point-biserial

correlation between Gender and the Depression T score. Compare the relevant aspects of

this question to the results you obtained in Exercise 7.46. (See “ The Relationship Between

and t” within Section 10.1.)

10.19 In Exercise 7.48 using Mireault.dat, we compared the responses of students who had lost

a parent and students who had not lost a parent in terms of their responses on the Global

Symptom Index T score (GSIT), among other variables. An alternative analysis would be

to use a clinically meaningful cutoff on the GSIT, classifying anyone over that score as a

clinical case (showing a clinically significant level of symptoms) and everyone below that

score as a noncase. Derogatis (1983) has suggested a score of 63 as the cutoff (e.g., if

GSIT .63 then ClinCase 5 1; else ClinCase 5 0).

a. Use any statistical package to create the variable of ClinCase, as defined by Derogatis.

Then cross-tabulate ClinCase against Group. Compute chi-square and Cramér’s C.

b. How does the answer to part (a) compare to the answers obtained in Chapter 7?

c. Why might we prefer this approach (looking at case versus noncase) over the proce-

dure adopted in Chapter 7?

(Hint: SAS will require Proc Freq; and SPSS will use CrossTabs. The appropriate manu-

als will help you set up the commands.)

10.20 Repeat the analysis shown in Exercise 10.19, but this time cross-tabulate ClinCase

against Gender.

a. Compare this answer with the results of Exercise 10.18.

b. How does this analysis differ from the one in Exercise 10.18 on roughly the same

question?

f

rpb

x^2

r^2

x^2

r^2