version 12.1 * Problem 1B clear all matrix input means = (79\14\15) matrix input sds = (9.4\2.7\5.6) matrix input corr = (1,.35,-.22\.35,1,-.90\-.22,-.90,1) corr2data opinion educ tv, n(200) means(means) sds(sds) corr(corr) sum corr corr, cov reg opinion educ tv, beta reg opinion tv, beta reg opinion educ, beta *Problem 2 use http://www3.nd.edu/~rwilliam/xsoc63993/statafiles/gender.dta, clear *** Wald Test reg income educ jobexp test educ = jobexp vce *** Incremental F test * Unconstrained model reg income educ jobexp est store unconstrained * Constrained model gen jobed = educ + jobexp reg income jobed est store constrained * Use Buis's -ftest- command ftest constrained unconstrained *** Likelihood Ratio Test lrtest constrained unconstrained, stats * Part 3 use http://www3.nd.edu/~rwilliam/xsoc63993/statafiles/gender.dta, clear * (a) Compositional differences ttest educ, by(female) ttest jobexp, by(female) ttest income, by(female) * (b) Differences in effects * Constrained model: No Gender differences reg income educ jobexp est store both * Unconstrained - Effects differ by gender reg income educ jobexp if female == 0 est store male reg income educ jobexp if female == 1 est store female *** Likelihood ratio lrtest (male female) both *** Wald test using suest quietly reg income educ jobexp if female == 0 est store male quietly reg income educ jobexp if female == 1 est store female suest male female test [male_mean = female_mean], constant coef * (d) What if tabstat income educ jobexp, by(female) columns(variables) * Margins quietly reg income educ jobexp if female == 1 margins, at (educ = 11.2222 jobexp = 14.11111) * Predict quietly reg income educ jobexp if female == 1 adjust educ jobexp if female == 0 * Adjust quietly reg income educ jobexp if female == 1 adjust educ jobexp if female == 0 * Margins using atmeans quietly reg income educ jobexp if female == 1 margins if female == 0, atmeans noesample * Margins using precise values sum educ if female == 0, meanonly scalar malemeaneduc = r(mean) sum jobexp if female == 0, meanonly scalar malemeanjobexp = r(mean) scalar list quietly reg income educ jobexp if female == 1 margins, at (educ = `=malemeaneduc' jobexp = `=malemeanjobexp')