Alternative predictors

Example 6.3 of Mize, Doan, and Long (2019)

Two models can differ in how the same construct is measured. Here support for same-sex relationships is predicted by sexual orientation measured two ways: by reported sexual behavior in one model and by self-identification in the other. mecompare puts the marginal effects of both measures in one table and allows a test of whether the measures give the same answer.

use "https://tdmize.github.io/data/data/gss_cme", clear

(gss_cme.dta | GSS 1972 - 2016 Weighted | 2018-07-10)

drop if missing(samesexB, sexident, sexbehav, college, woman, race, age, year)

(57,545 observations deleted)

. 

quietly logit samesexB i.sexbehav i.woman i.college c.age i.race i.year, vce(robust)
estimates store behavior

. 

quietly logit samesexB i.sexident i.woman i.college c.age i.race i.year, vce(robust)
estimates store identity

Each predictor appears in only one of the models, so the table reports each effect for the model that contains it:

mecompare i.sexbehav i.sexident, models(behavior identity)

Predicting: Pr(samesexB)

Marginal effects and cross-model differences (N_behavior=4921) (N_identity=4921)

                                 |  ME #   Estimate  Robust SE      P>|z|
---------------------------------+---------------------------------------
sexbehav                         |                                       
     Bisexual - Heterosexual     |                                       
                        behavior |     1     -0.097      0.023      0.000
                        identity |                                       
                      Difference |                                       
          Gay - Heterosexual     |                                       
                        behavior |     2     -0.362      0.038      0.000
                        identity |                                       
                      Difference |                                       
---------------------------------+---------------------------------------
sexident                         |                                       
     Bisexual - Heterosexual     |                                       
                        behavior |                                       
                        identity |     3     -0.274      0.041      0.000
                      Difference |                                       
          Gay - Heterosexual     |                                       
                        behavior |                                       
                        identity |     4     -0.428      0.034      0.000
                      Difference |                                       

Contrasts across the two measures are contrasts between rows of this table, referred to by the number in the # column. Is the bisexual versus heterosexual difference the same when sexual orientation is measured by behavior (row 1) as when it is measured by identity (row 3)? And is the gay versus heterosexual difference the same across the two measures (rows 2 and 4)? metest answers both:

metest 1 - 3

                                 |  estimate         se     pvalue 
---------------------------------+--------------------------------
 sexbehav_Bis~h - sexident_Bis~e |     0.177      0.043      0.000 

metest 2 - 4

                                 |  estimate         se     pvalue 
---------------------------------+--------------------------------
 sexbehav_Gay~r - sexident_Gay~y |     0.066      0.041      0.106 

Both marginal effects are negative, so a positive difference means the effect is smaller in magnitude when orientation is measured by behavior than when it is measured by identity.