We take another look at the frequency of political discussions (DISCPOL), first presented in Table 10.11 (based on GSS 2014). DISCPOL is presented with two independent variables—(1) respondent sex and (2) age (measured in categories as reported in corresponding tables). Interpret each measure of association. Which variable—age or sex—has a higher proportional reduction of error?

Table 10.11

Transcribed Image Text:

discpol HOW OFTEN R DISCUSS POLITICS RAge recoded Age Crosstabulation Count RAge recoded Age 1.00 18-29 2.00 30-39 3.00 40-49 4.00 50-59 Total discpol HOW OFTEN R DISCUSS POLITICS 1 OFTEN 2 SOMETIMES 3 20 12 17 16 26 71 3 RARELY 19 25 23 24 91 4 NEVER 18 31 15 14 78 Total 52 78 58 72 260 Symmetric Measures Asymptotic Approximate Standardized Error Approximate Šignificance Value Ordinal by Ordinal Gamma -.206 .070 -2.932 .003 N of Valid Cases a. Not assuming the null hypothesis. b. Using the asymptotic standard error assuming the null hypothesis. 260 discpol HOW OFTEN R DISCUSS POLITICS sex RESPONDENTS SEX Crosstabulation Count sex RESPONDENTS SEX I MALE 2 FEMALE Total discpol HOW OFTEN R DISCUSS POLITICS 1 OFTEN 2 SOMETIMES 3 RARELY 21 14 35 57 49 106 49 82 131 4 NEVER 32 64 96 Total 159 209 368 Directional Measures Asymptotic Standardized Error Аpprox mate Approximate Šignificance Value Nominal by Nominal Lambda Symmetric discpol HOW OFTÈN R DISCUSS POLITICS Dependent .058 .047 1.213 225 .034 .043 .778 .437 sex RESPONDENTS SEX Dependent .094 .071 1.266 .206 discpol HOW OFTEN R DISCUSS POUTICS Dependent Goodman and Kruskal tau .013 .007 .003 sex RESPONDENTS SEX Dependent 039 .020 .002 a. Not assuming the null hypothesis. b. Using the asymptotic standard error assuming the null hypothesis. C. Based on chi-square approximation