metest

Tests and contrasts of marginal effects

metest tests linear and nonlinear combinations of the marginal effects reported by mecompare – or of any results posted by margins, post. Rows are referred to by the number in the # column of the mecompare table, so a difference between two effects is metest 2 - 1, a ratio is metest 2 / 1, and a test that several effects are equal is metest 1 = 2 = 3. To refer to an actual number for a mathematical expression, use a leading # – e.g. to divide the 5th marginal effect by 2 use metest 5 / #2.

Sums, differences, ratios, products, and functions such as abs(), exp(), ln(), and min() are all allowed, as are equality tests with =, including several at once. Results can be accumulated into one table with add.

Installation

metest is part of the mecompare package and is installed with it:

net install mecompare, from("https://tdmize.github.io/data") replace

metest requires Stata 16 or later.

Where to start

Usage and examples runs every form of the command from the help file: single effects, differences, ratios and averages, scaling by a literal value, equality tests, accumulated tables, and use after commands other than mecompare.
Tests across three or more models compares one effect across four nested models.
The help file is on the site.
The mecompare pages use metest throughout, e.g. the effect sizes across predictors example and the interactions pages.

Basic use

sysuse nlsw88, clear

(NLSW, 1988 extract)

drop if missing(union)

(368 observations deleted)

.

quietly logit union i.collgrad i.south i.smsa c.age c.wage, vce(robust)
estimates store m

mecompare i.collgrad i.south i.smsa, models(m)

Predicting: Pr(union)

Marginal effects (N_m=1878)

                                 |  ME #   Estimate  Robust SE      P>|z|
---------------------------------+---------------------------------------
collgrad                         |                                       
    College gra - Not colleg     |                                       
                               m |     1      0.056      0.025      0.026
---------------------------------+---------------------------------------
south                            |                                       
           South - Not south     |                                       
                               m |     2     -0.114      0.020      0.000
---------------------------------+---------------------------------------
smsa                             |                                       
             SMSA - Not SMSA     |                                       
                               m |     3      0.031      0.022      0.161

Is the college gap in union membership the same size as the regional gap? And is the college gap larger than the city/non-city gap?

metest 1 - 2

                                 |  estimate         se     pvalue 
---------------------------------+--------------------------------
                collgrad - south |     0.170      0.034      0.000

metest 1 - 3

                                 |  estimate         se     pvalue 
---------------------------------+--------------------------------
                 collgrad - smsa |     0.025      0.034      0.465