sysuse auto * model selection, backward and foreward stepwise, pe(0.1): reg price weight trunk length mpg turn displacement gear_ratio headroom foreign rep78 // begin with empty model, forewards stepwise, pr(0.1): reg price weight trunk length mpg turn displacement gear_ratio headroom foreign rep78 // begin with full model backwards stepwise, pe(0.1): logit foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 // begin with empty model, forewards stepwise, pe(0.1): probit foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 // begin with empty model, forewards stepwise, pr(0.1): logit foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 // begin with full model backwards stepwise, pr(0.1): probit foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 // begin with full model backwards stepwise, pe(0.1): cloglog foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 // begin with empty model, backwards stepwise, pr(0.1): cloglog foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 // begin with full model backwards *compare models; the selected above model has lower AIC, so better probit foreign gear_ratio rep78 mpg estat ic est store m1 probit foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 est store m2 lrtest m1 m2, stats est table m1 m2 , stats(r2_p aic bic) *repeat using logit model *compare models; the selected above model has lower AIC, so better logit foreign gear_ratio rep78 mpg estat ic est store m3 logit foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 est store m4 lrtest m3 m4, stats est table m3 m4 , stats(r2_p aic bic) est table m1 m2 m3 m4 , stats(r2_p aic bic) // compare all four models cloglog foreign displacement gear_ratio rep78 mpg estat ic est store m5 cloglog foreign weight trunk length mpg turn displacement gear_ratio price headroom rep78 est store m6 lrtest m5 m6, stats est table m5 m6 , stats(r2_p aic bic) est table m1 m2 m3 m4 m5 m6, stats(r2_p aic bic) // compare all 6 models // logit with selected variables > probit with selected variables > cloglog with selected variables. // conclusion: logit with selected variables is the best model clear