Examples from the help file
The examples at the end of help mecompare, run in full
These are the examples from the end of the help file, in the same order, with their output. They use Stata’s nlsw88 data; every line can be pasted into Stata as is.
Fit and store the models, then compare marginal effects
Factor syntax is required for the regression models; it is optional for mecompare.
sysuse nlsw88, clear(NLSW, 1988 extract)
. logit union i.married age i.race hours i.collgrad, vce(robust)Iteration 0: Log pseudolikelihood = -1046.3424
Iteration 1: Log pseudolikelihood = -1026.8839
Iteration 2: Log pseudolikelihood = -1026.721
Iteration 3: Log pseudolikelihood = -1026.721
Logistic regression Number of obs = 1,877
Wald chi2(6) = 38.14
Prob > chi2 = 0.0000
Log pseudolikelihood = -1026.721 Pseudo R2 = 0.0188
-------------------------------------------------------------------------------
| Robust
union | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
married |
Married | -0.137 0.116 -1.175 0.240 -0.364 0.091
age | 0.014 0.018 0.764 0.445 -0.021 0.049
|
race |
Black | 0.428 0.122 3.512 0.000 0.189 0.667
Other | 0.521 0.453 1.149 0.251 -0.368 1.409
|
hours | 0.010 0.006 1.836 0.066 -0.001 0.021
|
collgrad |
College grad | 0.526 0.120 4.368 0.000 0.290 0.762
_cons | -2.231 0.757 -2.948 0.003 -3.714 -0.748
-------------------------------------------------------------------------------est store basemod. logit union i.married age i.race hours i.collgrad wage, vce(robust)Iteration 0: Log pseudolikelihood = -1046.3424
Iteration 1: Log pseudolikelihood = -1013.2125
Iteration 2: Log pseudolikelihood = -1012.7232
Iteration 3: Log pseudolikelihood = -1012.7232
Logistic regression Number of obs = 1,877
Wald chi2(7) = 65.25
Prob > chi2 = 0.0000
Log pseudolikelihood = -1012.7232 Pseudo R2 = 0.0321
-------------------------------------------------------------------------------
| Robust
union | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
married |
Married | -0.099 0.117 -0.846 0.398 -0.329 0.131
age | 0.014 0.018 0.780 0.435 -0.021 0.050
|
race |
Black | 0.502 0.123 4.083 0.000 0.261 0.742
Other | 0.474 0.490 0.966 0.334 -0.487 1.435
|
hours | 0.008 0.006 1.342 0.180 -0.004 0.019
|
collgrad |
College grad | 0.282 0.132 2.132 0.033 0.023 0.542
wage | 0.071 0.013 5.376 0.000 0.045 0.097
_cons | -2.698 0.774 -3.486 0.000 -4.215 -1.181
-------------------------------------------------------------------------------est store medmod. mecompare age collgrad race hours, models(basemod medmod)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod | 1 0.002 0.003 0.445
medmod | 2 0.003 0.003 0.435
Difference | 3 -0.000 0.000 0.906
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
basemod | 4 0.102 0.024 0.000
medmod | 5 0.052 0.025 0.039
Difference | 6 0.050 0.009 0.000
---------------------------------+---------------------------------------
race |
Black - White |
basemod | 7 0.081 0.024 0.001
medmod | 8 0.094 0.024 0.000
Difference | 9 -0.013 0.003 0.000
Other - White |
basemod | 10 0.101 0.097 0.299
medmod | 11 0.088 0.101 0.380
Difference | 12 0.012 0.014 0.391
---------------------------------+---------------------------------------
hours + 1 (centered) |
basemod | 13 0.002 0.001 0.066
medmod | 14 0.001 0.001 0.179
Difference | 15 0.000 0.000 0.009Amount of change for continuous variables
mecompare age collgrad race hours, models(basemod medmod) amount(sd)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + SD (centered) |
basemod | 1 0.008 0.010 0.445
medmod | 2 0.008 0.010 0.435
Difference | 3 -0.000 0.001 0.906
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
basemod | 4 0.102 0.024 0.000
medmod | 5 0.052 0.025 0.039
Difference | 6 0.050 0.009 0.000
---------------------------------+---------------------------------------
race |
Black - White |
basemod | 7 0.081 0.024 0.001
medmod | 8 0.094 0.024 0.000
Difference | 9 -0.013 0.003 0.000
Other - White |
basemod | 10 0.101 0.097 0.299
medmod | 11 0.088 0.101 0.380
Difference | 12 0.012 0.014 0.391
---------------------------------+---------------------------------------
hours + SD (centered) |
basemod | 13 0.018 0.010 0.066
medmod | 14 0.014 0.010 0.179
Difference | 15 0.005 0.002 0.009mecompare age collgrad race hours, models(basemod medmod) amount(sd 10)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + SD (centered) |
basemod | 1 0.008 0.010 0.445
medmod | 2 0.008 0.010 0.435
Difference | 3 -0.000 0.001 0.906
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
basemod | 4 0.102 0.024 0.000
medmod | 5 0.052 0.025 0.039
Difference | 6 0.050 0.009 0.000
---------------------------------+---------------------------------------
race |
Black - White |
basemod | 7 0.081 0.024 0.001
medmod | 8 0.094 0.024 0.000
Difference | 9 -0.013 0.003 0.000
Other - White |
basemod | 10 0.101 0.097 0.299
medmod | 11 0.088 0.101 0.380
Difference | 12 0.012 0.014 0.391
---------------------------------+---------------------------------------
hours + 10 (centered) |
basemod | 13 0.019 0.010 0.066
medmod | 14 0.014 0.010 0.179
Difference | 15 0.005 0.002 0.009amount(rate) gives the instantaneous rate of change; amount(dydx) and amount(slope) are synonyms.
mecompare age hours, models(basemod medmod) amount(rate)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age(rate) |
basemod | 1 0.002 0.003 0.445
medmod | 2 0.003 0.003 0.435
Difference | 3 -0.000 0.000 0.906
---------------------------------+---------------------------------------
hours(rate) |
basemod | 4 0.002 0.001 0.066
medmod | 5 0.001 0.001 0.179
Difference | 6 0.000 0.000 0.009mecompare age hours, models(basemod medmod) amount(range)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age(min-max) |
basemod | 1 0.030 0.039 0.447
medmod | 2 0.030 0.039 0.437
Difference | 3 -0.001 0.005 0.906
---------------------------------+---------------------------------------
hours(min-max) |
basemod | 4 0.148 0.081 0.068
medmod | 5 0.111 0.083 0.183
Difference | 6 0.037 0.014 0.009Labeling the models
mecompare age collgrad race hours, models(basemod medmod) mod1name(Base Model) mod2name(Mediation
Model)Predicting: Pr(union)
Marginal effects and cross-model differences (N_Base Model=1877) (N_Mediation Model=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
Base Model | 1 0.002 0.003 0.445
Mediation Model | 2 0.003 0.003 0.435
Difference | 3 -0.000 0.000 0.906
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
Base Model | 4 0.102 0.024 0.000
Mediation Model | 5 0.052 0.025 0.039
Difference | 6 0.050 0.009 0.000
---------------------------------+---------------------------------------
race |
Black - White |
Base Model | 7 0.081 0.024 0.001
Mediation Model | 8 0.094 0.024 0.000
Difference | 9 -0.013 0.003 0.000
Other - White |
Base Model | 10 0.101 0.097 0.299
Mediation Model | 11 0.088 0.101 0.380
Difference | 12 0.012 0.014 0.391
---------------------------------+---------------------------------------
hours + 1 (centered) |
Base Model | 13 0.002 0.001 0.066
Mediation Model | 14 0.001 0.001 0.179
Difference | 15 0.000 0.000 0.009Effects within levels of a variable (by, over)
mecompare age race, models(basemod medmod) over(collgrad)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod Not college grad | 1 0.002 0.003 0.445
basemod College grad | 2 0.003 0.004 0.445
medmod Not college grad | 3 0.002 0.003 0.435
medmod College grad | 4 0.003 0.004 0.435
Difference Not college grad | 5 -0.000 0.000 0.884
Difference College grad | 6 -0.000 0.000 0.960
---------------------------------+---------------------------------------
race |
Black - White |
basemod Not college grad | 7 0.076 0.023 0.001
basemod College grad | 8 0.096 0.028 0.001
medmod Not college grad | 9 0.089 0.023 0.000
medmod College grad | 10 0.110 0.028 0.000
Difference Not college grad | 11 -0.013 0.003 0.000
Difference College grad | 12 -0.014 0.004 0.000
Other - White |
basemod Not college grad | 13 0.095 0.093 0.307
basemod College grad | 14 0.118 0.109 0.279
medmod Not college grad | 15 0.083 0.096 0.387
medmod College grad | 16 0.104 0.114 0.362
Difference Not college grad | 17 0.012 0.014 0.400
Difference College grad | 18 0.015 0.016 0.372mecompare age race, models(basemod medmod) by(collgrad)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod Not college grad | 1 0.002 0.003 0.445
basemod College grad | 2 0.003 0.004 0.445
medmod Not college grad | 3 0.002 0.003 0.435
medmod College grad | 4 0.003 0.004 0.435
Difference Not college grad | 5 -0.000 0.000 0.751
Difference College grad | 6 0.000 0.001 0.777
---------------------------------+---------------------------------------
race |
Black - White |
basemod Not college grad | 7 0.077 0.023 0.001
basemod College grad | 8 0.096 0.028 0.001
medmod Not college grad | 9 0.092 0.023 0.000
medmod College grad | 10 0.103 0.027 0.000
Difference Not college grad | 11 -0.015 0.004 0.000
Difference College grad | 12 -0.008 0.004 0.039
Other - White |
basemod Not college grad | 13 0.095 0.093 0.307
basemod College grad | 14 0.118 0.109 0.280
medmod Not college grad | 15 0.086 0.099 0.384
medmod College grad | 16 0.097 0.109 0.372
Difference Not college grad | 17 0.009 0.015 0.522
Difference College grad | 18 0.021 0.016 0.191mecompare age race, models(basemod medmod) by(collgrad married)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod Not college grad Sin | 1 0.002 0.003 0.444
basemod Not college grad Mar | 2 0.002 0.003 0.445
basemod College grad Single | 3 0.003 0.004 0.445
basemod College grad Married | 4 0.003 0.004 0.446
medmod Not college grad Sing | 5 0.003 0.003 0.434
medmod Not college grad Marr | 6 0.002 0.003 0.435
medmod College grad Single | 7 0.003 0.004 0.435
medmod College grad Married | 8 0.003 0.004 0.435
Difference Not college grad | 9 -0.000 0.000 0.816
Difference Not college grad | 10 -0.000 0.000 0.716
Difference College grad Sing | 11 0.000 0.001 0.763
Difference College grad Marr | 12 0.000 0.001 0.783
---------------------------------+---------------------------------------
race |
Black - White |
basemod Not college grad Sin | 13 0.080 0.023 0.001
basemod Not college grad Mar | 14 0.075 0.023 0.001
basemod College grad Single | 15 0.098 0.029 0.001
basemod College grad Married | 16 0.094 0.028 0.001
medmod Not college grad Sing | 17 0.095 0.024 0.000
medmod Not college grad Marr | 18 0.090 0.023 0.000
medmod College grad Single | 19 0.106 0.027 0.000
medmod College grad Married | 20 0.102 0.027 0.000
Difference Not college grad | 21 -0.014 0.004 0.000
Difference Not college grad | 22 -0.016 0.004 0.000
Difference College grad Sing | 23 -0.008 0.004 0.056
Difference College grad Marr | 24 -0.008 0.004 0.035
Other - White |
basemod Not college grad Sin | 25 0.100 0.097 0.304
basemod Not college grad Mar | 26 0.093 0.092 0.309
basemod College grad Single | 27 0.121 0.111 0.276
basemod College grad Married | 28 0.116 0.108 0.283
medmod Not college grad Sing | 29 0.089 0.101 0.382
medmod Not college grad Marr | 30 0.085 0.097 0.385
medmod College grad Single | 31 0.100 0.111 0.370
medmod College grad Married | 32 0.096 0.108 0.374
Difference Not college grad | 33 0.011 0.015 0.459
Difference Not college grad | 34 0.008 0.015 0.561
Difference College grad Sing | 35 0.021 0.016 0.187
Difference College grad Marr | 36 0.020 0.016 0.195Whether an effect differs across the levels (a test of interaction), jointly and for one pair:
mecompare age, models(basemod) by(race)Predicting: Pr(union)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
White | 1 0.002 0.003 0.444
Black | 2 0.003 0.004 0.446
Other | 3 0.003 0.004 0.447metest 1 = 2 = 3Tests of equality
| chi2 df pvalue
---------------------------------+--------------------------------
age_ra~1 = age_ra~2 = age_ra~3 | 0.547 2.000 0.761 metest 1 - 2 | estimate se pvalue
---------------------------------+--------------------------------
age_race_1 - age_race_2 | -0.001 0.001 0.461 For a nominal variable, one equality per contrast:
mecompare race, models(basemod) by(collgrad)Predicting: Pr(union)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
race |
Black - White |
Not college grad | 1 0.077 0.023 0.001
College grad | 2 0.096 0.028 0.001
Other - White |
Not college grad | 3 0.095 0.093 0.307
College grad | 4 0.118 0.109 0.280metest (1 = 2) (3 = 4)Tests of equality
| chi2 df pvalue
---------------------------------+--------------------------------
(rac~0 = rac~1)(rac~0 = rac~1) | 7.978 2.000 0.019 Values of the focal variables and covariates, including lists of values
mecompare age race, models(basemod medmod) atmeansPredicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod | 1 0.002 0.003 0.445
medmod | 2 0.003 0.003 0.435
Difference | 3 -0.000 0.000 0.865
---------------------------------+---------------------------------------
race |
Black - White |
basemod | 4 0.082 0.024 0.001
medmod | 5 0.096 0.025 0.000
Difference | 6 -0.014 0.004 0.000
Other - White |
basemod | 7 0.102 0.098 0.301
medmod | 8 0.090 0.103 0.383
Difference | 9 0.012 0.015 0.438mecompare age, models(basemod) start(age=(35 40 45))Predicting: Pr(union)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
at 35 | 1 0.002 0.003 0.431
at 40 | 2 0.002 0.003 0.447
at 45 | 3 0.003 0.003 0.461mecompare age, models(basemod) start(age=35) end(age=40)Predicting: Pr(union)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age 35 to 40 |
basemod | 1 0.012 0.016 0.439mecompare age, models(basemod) start(age=(35 40)) end(age=(40 45))Predicting: Pr(union)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age |
35 to 40 | 1 0.012 0.016 0.439
40 to 45 | 2 0.013 0.017 0.454mecompare age race, models(basemod medmod) covariates(hours=(20 40 60))Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod hours=20 | 1 0.002 0.003 0.445
basemod hours=40 | 2 0.003 0.003 0.445
basemod hours=60 | 3 0.003 0.004 0.445
medmod hours=20 | 4 0.002 0.003 0.435
medmod hours=40 | 5 0.003 0.003 0.435
medmod hours=60 | 6 0.003 0.004 0.435
Difference hours=20 | 7 -0.000 0.000 0.792
Difference hours=40 | 8 -0.000 0.000 0.922
Difference hours=60 | 9 0.000 0.000 0.973
---------------------------------+---------------------------------------
race |
Black - White |
basemod hours=20 | 10 0.074 0.022 0.001
basemod hours=40 | 11 0.082 0.024 0.001
basemod hours=60 | 12 0.090 0.027 0.001
medmod hours=20 | 13 0.088 0.023 0.000
medmod hours=40 | 14 0.095 0.024 0.000
medmod hours=60 | 15 0.101 0.026 0.000
Difference hours=20 | 16 -0.014 0.004 0.000
Difference hours=40 | 17 -0.013 0.003 0.000
Difference hours=60 | 18 -0.012 0.004 0.001
Other - White |
basemod hours=20 | 19 0.093 0.091 0.309
basemod hours=40 | 20 0.102 0.098 0.298
basemod hours=60 | 21 0.111 0.104 0.288
medmod hours=20 | 22 0.083 0.096 0.387
medmod hours=40 | 23 0.089 0.102 0.379
medmod hours=60 | 24 0.095 0.107 0.373
Difference hours=20 | 25 0.010 0.014 0.483
Difference hours=40 | 26 0.013 0.015 0.379
Difference hours=60 | 27 0.015 0.015 0.317Summary measures for nominal variables
mecompare race, models(basemod medmod) pwcomparePredicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
race |
Black - White |
basemod | 1 0.081 0.024 0.001
medmod | 2 0.094 0.024 0.000
Difference | 3 -0.013 0.003 0.000
Other - White |
basemod | 4 0.101 0.097 0.299
medmod | 5 0.088 0.101 0.380
Difference | 6 0.012 0.014 0.391
Other - Black |
basemod | 7 0.020 0.099 0.843
medmod | 8 -0.006 0.102 0.955
Difference | 9 0.025 0.015 0.088mecompare race, models(basemod) meineq(weighted)Predicting: Pr(union)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
race |
ME Inequality | 1 0.080 0.050 0.113
Black - White |
basemod | 2 0.081 0.024 0.001
Other - White |
basemod | 3 0.101 0.097 0.299mecompare race, models(basemod medmod) meineq(all)Predicting: Pr(union)
Marginal effects and cross-model differences (N_basemod=1877) (N_medmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
race |
ME Inequality |
basemod | 1 0.080 0.050 0.113
medmod | 2 0.080 0.028 0.004
Difference | 3 0.000 0.027 0.998
Unwgt ME Ineq. |
basemod | 4 0.067 0.065 0.299
medmod | 5 0.063 0.016 0.000
Difference | 6 0.004 0.066 0.946
Black - White |
basemod | 7 0.081 0.024 0.001
medmod | 8 0.094 0.024 0.000
Difference | 9 -0.013 0.003 0.000
Other - White |
basemod | 10 0.101 0.097 0.299
medmod | 11 0.088 0.101 0.380
Difference | 12 0.012 0.014 0.391Total ME, which sums effects across outcome categories
For example, with an ordinal outcome:
egen hours_ord = cut(hours), at(0 30 40 50 81)(4 missing values generated)ologit hours_ord c.age i.married i.race, vce(robust)Iteration 0: Log pseudolikelihood = -2558.0934
Iteration 1: Log pseudolikelihood = -2535.2485
Iteration 2: Log pseudolikelihood = -2535.1689
Iteration 3: Log pseudolikelihood = -2535.1689
Ordered logistic regression Number of obs = 2,242
Wald chi2(4) = 46.67
Prob > chi2 = 0.0000
Log pseudolikelihood = -2535.1689 Pseudo R2 = 0.0090
------------------------------------------------------------------------------
| Robust
hours_ord | Coefficient std. err. z P>|z| [95% conf. interval]
-------------+----------------------------------------------------------------
age | -0.031 0.013 -2.352 0.019 -0.057 -0.005
|
married |
Married | -0.523 0.085 -6.150 0.000 -0.690 -0.357
|
race |
Black | 0.080 0.084 0.947 0.344 -0.086 0.246
Other | 0.291 0.429 0.679 0.497 -0.549 1.131
-------------+----------------------------------------------------------------
/cut1 | -3.193 0.528 -4.228 -2.159
/cut2 | -2.207 0.527 -3.240 -1.174
/cut3 | 0.793 0.526 -0.237 1.823
------------------------------------------------------------------------------mecompare age married race, totalmePredicting: Pr(hours_ord)
Marginal effects (N_m1=2242)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
Total ME | 1 0.007 0.003 0.018
Outcome 0 - m1 | 2 0.004 0.002 0.019
Outcome 30 - m1 | 3 0.003 0.001 0.018
Outcome 40 - m1 | 4 -0.004 0.002 0.019
Outcome 50 - m1 | 5 -0.003 0.001 0.019
---------------------------------+---------------------------------------
married |
Married - Single |
Total ME | 6 0.114 0.018 0.000
Outcome 0 - m1 | 7 0.067 0.011 0.000
Outcome 30 - m1 | 8 0.047 0.008 0.000
Outcome 40 - m1 | 9 -0.068 0.011 0.000
Outcome 50 - m1 | 10 -0.045 0.008 0.000
---------------------------------+---------------------------------------
race |
Total ME Ineq. | 11 0.038 0.045 0.400
Black - White |
Outcome 0 - m1 | 12 -0.011 0.011 0.340
Outcome 30 - m1 | 13 -0.007 0.007 0.345
Outcome 40 - m1 | 14 0.011 0.011 0.340
Outcome 50 - m1 | 15 0.007 0.007 0.346
Other - White |
Outcome 0 - m1 | 16 -0.036 0.048 0.452
Outcome 30 - m1 | 17 -0.025 0.038 0.502
Outcome 40 - m1 | 18 0.035 0.043 0.410
Outcome 50 - m1 | 19 0.026 0.043 0.542Comparing models fit over distinct groups
logit union age hours i.collgrad if south==1, vce(robust)Iteration 0: Log pseudolikelihood = -364.3358
Iteration 1: Log pseudolikelihood = -361.88094
Iteration 2: Log pseudolikelihood = -361.86467
Iteration 3: Log pseudolikelihood = -361.86467
Logistic regression Number of obs = 798
Wald chi2(3) = 4.95
Prob > chi2 = 0.1758
Log pseudolikelihood = -361.86467 Pseudo R2 = 0.0068
-------------------------------------------------------------------------------
| Robust
union | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
age | -0.033 0.030 -1.104 0.270 -0.091 0.026
hours | 0.017 0.010 1.681 0.093 -0.003 0.037
|
collgrad |
College grad | 0.183 0.212 0.861 0.389 -0.233 0.598
_cons | -1.012 1.228 -0.824 0.410 -3.418 1.394
-------------------------------------------------------------------------------est store msouth
logit union age hours i.collgrad if south==0, vce(robust)Iteration 0: Log pseudolikelihood = -660.22034
Iteration 1: Log pseudolikelihood = -647.21277
Iteration 2: Log pseudolikelihood = -647.13776
Iteration 3: Log pseudolikelihood = -647.13776
Logistic regression Number of obs = 1,079
Wald chi2(3) = 26.00
Prob > chi2 = 0.0000
Log pseudolikelihood = -647.13776 Pseudo R2 = 0.0198
-------------------------------------------------------------------------------
| Robust
union | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
age | 0.032 0.022 1.445 0.148 -0.012 0.076
hours | 0.015 0.006 2.344 0.019 0.002 0.027
|
collgrad |
College grad | 0.613 0.148 4.152 0.000 0.324 0.902
_cons | -2.838 0.923 -3.075 0.002 -4.647 -1.029
-------------------------------------------------------------------------------est store mnonsouth. mecompare age hours collgrad, models(msouth mnonsouth) groups groupnames(South NonSouth)Predicting: Pr(union)
Marginal effects and cross-model differences (N_South=798) (N_NonSouth=1079)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
South | 1 -0.005 0.004 0.269
NonSouth | 2 0.007 0.005 0.147
Difference | 3 -0.011 0.006 0.069
---------------------------------+---------------------------------------
hours + 1 (centered) |
South | 4 0.002 0.001 0.092
NonSouth | 5 0.003 0.001 0.018
Difference | 6 -0.001 0.002 0.733
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
South | 7 0.026 0.032 0.402
NonSouth | 8 0.134 0.033 0.000
Difference | 9 -0.108 0.046 0.019Cross-family comparison (same number of predictions)
logit union i.married age i.race hours i.collgrad, vce(robust)Iteration 0: Log pseudolikelihood = -1046.3424
Iteration 1: Log pseudolikelihood = -1026.8839
Iteration 2: Log pseudolikelihood = -1026.721
Iteration 3: Log pseudolikelihood = -1026.721
Logistic regression Number of obs = 1,877
Wald chi2(6) = 38.14
Prob > chi2 = 0.0000
Log pseudolikelihood = -1026.721 Pseudo R2 = 0.0188
-------------------------------------------------------------------------------
| Robust
union | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
married |
Married | -0.137 0.116 -1.175 0.240 -0.364 0.091
age | 0.014 0.018 0.764 0.445 -0.021 0.049
|
race |
Black | 0.428 0.122 3.512 0.000 0.189 0.667
Other | 0.521 0.453 1.149 0.251 -0.368 1.409
|
hours | 0.010 0.006 1.836 0.066 -0.001 0.021
|
collgrad |
College grad | 0.526 0.120 4.368 0.000 0.290 0.762
_cons | -2.231 0.757 -2.948 0.003 -3.714 -0.748
-------------------------------------------------------------------------------est store logitmod
regress union i.married age i.race hours i.collgrad, vce(robust)Linear regression Number of obs = 1,877
F(6, 1870) = 6.28
Prob > F = 0.0000
R-squared = 0.0212
Root MSE = .42665
-------------------------------------------------------------------------------
| Robust
union | Coefficient std. err. t P>|t| [95% conf. interval]
--------------+----------------------------------------------------------------
married |
Married | -0.025 0.022 -1.156 0.248 -0.068 0.018
age | 0.002 0.003 0.755 0.451 -0.004 0.009
|
race |
Black | 0.080 0.024 3.366 0.001 0.034 0.127
Other | 0.102 0.098 1.043 0.297 -0.090 0.295
|
hours | 0.002 0.001 1.862 0.063 -0.000 0.004
|
collgrad |
College grad | 0.102 0.024 4.181 0.000 0.054 0.149
_cons | 0.052 0.136 0.383 0.702 -0.214 0.318
-------------------------------------------------------------------------------est store lpmmod. mecompare age race, models(logitmod lpmmod)Predicting: Pr(union) (logitmod), Linear prediction (lpmmod)
Marginal effects and cross-model differences (N_logitmod=1877) (N_lpmmod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
logitmod | 1 0.002 0.003 0.445
lpmmod | 2 0.002 0.003 0.450
Difference | 3 0.000 0.000 0.522
---------------------------------+---------------------------------------
race |
Black - White |
logitmod | 4 0.081 0.024 0.001
lpmmod | 5 0.080 0.024 0.001
Difference | 6 0.001 0.001 0.195
Other - White |
logitmod | 7 0.101 0.097 0.299
lpmmod | 8 0.102 0.098 0.296
Difference | 9 -0.001 0.002 0.550Choosing the prediction
With one model, any prediction margins allows; a count model offers several.
nbreg hours c.age i.collgrad i.race, vce(robust)Fitting Poisson model:
Iteration 0: Log pseudolikelihood = -9931.8038
Iteration 1: Log pseudolikelihood = -9931.8038
Fitting constant-only model:
Iteration 0: Log pseudolikelihood = -10380.709
Iteration 1: Log pseudolikelihood = -9009.4719
Iteration 2: Log pseudolikelihood = -8762.6327
Iteration 3: Log pseudolikelihood = -8756.0008
Iteration 4: Log pseudolikelihood = -8755.9996
Iteration 5: Log pseudolikelihood = -8755.9996
Fitting full model:
Iteration 0: Log pseudolikelihood = -8746.2573
Iteration 1: Log pseudolikelihood = -8746.2051
Iteration 2: Log pseudolikelihood = -8746.2051
Negative binomial regression Number of obs = 2,242
Wald chi2(4) = 26.60
Dispersion: mean Prob > chi2 = 0.0000
Log pseudolikelihood = -8746.2051 Pseudo R2 = 0.0011
-------------------------------------------------------------------------------
| Robust
hours | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
age | -0.002 0.002 -1.107 0.268 -0.006 0.002
|
collgrad |
College grad | 0.059 0.015 4.007 0.000 0.030 0.088
|
race |
Black | 0.036 0.011 3.208 0.001 0.014 0.058
Other | -0.009 0.061 -0.151 0.880 -0.129 0.110
|
_cons | 3.678 0.076 48.202 0.000 3.528 3.827
--------------+----------------------------------------------------------------
/lnalpha | -2.609 0.084 -2.773 -2.445
--------------+----------------------------------------------------------------
alpha | 0.074 0.006 0.062 0.087
-------------------------------------------------------------------------------est store cntbase. mecompare age collgrad race, models(cntbase)Predicting: Predicted number of events
Marginal effects (N_cntbase=2242)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
cntbase | 1 -0.080 0.072 0.268
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
cntbase | 2 2.228 0.566 0.000
---------------------------------+---------------------------------------
race |
Black - White |
cntbase | 3 1.360 0.425 0.001
Other - White |
cntbase | 4 -0.338 2.227 0.879mecompare age collgrad race, models(cntbase) predict(n)Predicting: Predicted number of events
Marginal effects (N_cntbase=2242)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
cntbase | 1 -0.080 0.072 0.268
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
cntbase | 2 2.228 0.566 0.000
---------------------------------+---------------------------------------
race |
Black - White |
cntbase | 3 1.360 0.425 0.001
Other - White |
cntbase | 4 -0.338 2.227 0.879mecompare age collgrad race, models(cntbase) predict(ir)Predicting: Predicted incidence rate
Marginal effects (N_cntbase=2242)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
cntbase | 1 -0.080 0.072 0.268
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
cntbase | 2 2.228 0.566 0.000
---------------------------------+---------------------------------------
race |
Black - White |
cntbase | 3 1.360 0.425 0.001
Other - White |
cntbase | 4 -0.338 2.227 0.879mecompare age collgrad race, models(cntbase) predict(pr(0))Predicting: Pr(hours=0)
Marginal effects (N_cntbase=2242)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
cntbase | 1 0.000 0.000 0.356
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
cntbase | 2 -0.000 0.000 0.138
---------------------------------+---------------------------------------
race |
Black - White |
cntbase | 3 -0.000 0.000 0.189
Other - White |
cntbase | 4 0.000 0.000 0.885With two or more models, any prediction returning one quantity per model:
nbreg hours c.age i.collgrad i.race i.married, vce(robust)Fitting Poisson model:
Iteration 0: Log pseudolikelihood = -9871.3821
Iteration 1: Log pseudolikelihood = -9871.3821
Fitting constant-only model:
Iteration 0: Log pseudolikelihood = -10380.709
Iteration 1: Log pseudolikelihood = -9009.4719
Iteration 2: Log pseudolikelihood = -8762.6327
Iteration 3: Log pseudolikelihood = -8756.0008
Iteration 4: Log pseudolikelihood = -8755.9996
Iteration 5: Log pseudolikelihood = -8755.9996
Fitting full model:
Iteration 0: Log pseudolikelihood = -8730.3403
Iteration 1: Log pseudolikelihood = -8729.9693
Iteration 2: Log pseudolikelihood = -8729.9693
Negative binomial regression Number of obs = 2,242
Wald chi2(5) = 77.10
Dispersion: mean Prob > chi2 = 0.0000
Log pseudolikelihood = -8729.9693 Pseudo R2 = 0.0030
-------------------------------------------------------------------------------
| Robust
hours | Coefficient std. err. z P>|z| [95% conf. interval]
--------------+----------------------------------------------------------------
age | -0.003 0.002 -1.319 0.187 -0.006 0.001
|
collgrad |
College grad | 0.058 0.015 3.931 0.000 0.029 0.086
|
race |
Black | 0.019 0.011 1.630 0.103 -0.004 0.041
Other | -0.009 0.061 -0.141 0.888 -0.129 0.112
|
married |
Married | -0.081 0.012 -6.988 0.000 -0.103 -0.058
_cons | 3.749 0.076 49.502 0.000 3.600 3.897
--------------+----------------------------------------------------------------
/lnalpha | -2.632 0.084 -2.797 -2.467
--------------+----------------------------------------------------------------
alpha | 0.072 0.006 0.061 0.085
-------------------------------------------------------------------------------est store cntmed. mecompare age collgrad race, models(cntbase cntmed) predict(n)Predicting: Predicted mean of hours
Marginal effects and cross-model differences (N_cntbase=2242) (N_cntmed=2242)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
cntbase | 1 -0.080 0.072 0.268
cntmed | 2 -0.095 0.072 0.187
Difference | 3 0.015 0.010 0.156
---------------------------------+---------------------------------------
collgrad |
College gra - Not colleg |
cntbase | 4 2.228 0.566 0.000
cntmed | 5 2.174 0.562 0.000
Difference | 6 0.054 0.076 0.475
---------------------------------+---------------------------------------
race |
Black - White |
cntbase | 7 1.360 0.425 0.001
cntmed | 8 0.695 0.427 0.104
Difference | 9 0.665 0.116 0.000
Other - White |
cntbase | 10 -0.338 2.227 0.879
cntmed | 11 -0.319 2.255 0.888
Difference | 12 -0.019 0.277 0.945Passing an option to margins, e.g. effects in percentage points
mecompare age race, models(basemod) marginsopt(expression(100*predict(pr)))Predicting: 100*predict(pr)
Marginal effects (N_basemod=1877)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
age + 1 (centered) |
basemod | 1 0.247 0.324 0.445
---------------------------------+---------------------------------------
race |
Black - White |
basemod | 2 8.121 2.398 0.001
Other - White |
basemod | 3 10.086 9.715 0.299Four nested models and custom comparisons with metest
use https://tdmize.github.io/data/data/gss_cme, clear(gss_cme.dta | GSS 1972 - 2016 Weighted | 2018-07-10)drop if year < 2000(38,116 observations deleted)drop if employed != 1(9,556 observations deleted)drop if missing(vhappy, college, wages, occprest, age, married, parent, woman, conserv, reltrad)(5,578 observations deleted)
. logit vhappy i.college, vce(robust)Iteration 0: Log pseudolikelihood = -5696.026
Iteration 1: Log pseudolikelihood = -5671.3965
Iteration 2: Log pseudolikelihood = -5671.3676
Iteration 3: Log pseudolikelihood = -5671.3676
Logistic regression Number of obs = 9,216
Wald chi2(1) = 49.71
Prob > chi2 = 0.0000
Log pseudolikelihood = -5671.3676 Pseudo R2 = 0.0043
---------------------------------------------------------------------------------
| Robust
vhappy | Coefficient std. err. z P>|z| [95% conf. interval]
----------------+----------------------------------------------------------------
college |
College Degree | 0.331 0.047 7.051 0.000 0.239 0.423
_cons | -0.923 0.028 -32.493 0.000 -0.979 -0.867
---------------------------------------------------------------------------------estimates store m1
logit vhappy i.college i.married i.parent i.woman i.conserv i.reltrad i.year c.age##c.age, vce(rob
ust)Iteration 0: Log pseudolikelihood = -5696.026
Iteration 1: Log pseudolikelihood = -5415.347
Iteration 2: Log pseudolikelihood = -5412.2188
Iteration 3: Log pseudolikelihood = -5412.2175
Iteration 4: Log pseudolikelihood = -5412.2175
Logistic regression Number of obs = 9,216
Wald chi2(21) = 537.23
Prob > chi2 = 0.0000
Log pseudolikelihood = -5412.2175 Pseudo R2 = 0.0498
-----------------------------------------------------------------------------------
| Robust
vhappy | Coefficient std. err. z P>|z| [95% conf. interval]
------------------+----------------------------------------------------------------
college |
College Degree | 0.293 0.051 5.770 0.000 0.194 0.393
|
married |
Married | 0.997 0.052 19.176 0.000 0.895 1.099
|
parent |
Parent | -0.063 0.059 -1.059 0.290 -0.179 0.053
|
woman |
Women | 0.084 0.047 1.779 0.075 -0.009 0.177
|
conserv |
Conservative | 0.195 0.050 3.880 0.000 0.097 0.294
|
reltrad |
Mainline Prot | -0.095 0.077 -1.230 0.219 -0.245 0.056
Black Protestant | -0.168 0.100 -1.673 0.094 -0.365 0.029
Catholic | -0.133 0.065 -2.037 0.042 -0.260 -0.005
Jewish | -0.596 0.194 -3.079 0.002 -0.976 -0.217
Other Faith | -0.283 0.111 -2.548 0.011 -0.500 -0.065
Nonaffiliated | -0.258 0.073 -3.517 0.000 -0.402 -0.114
|
year |
2002 | 0.056 0.102 0.552 0.581 -0.143 0.256
2004 | -0.072 0.105 -0.686 0.493 -0.279 0.134
2006 | -0.147 0.084 -1.739 0.082 -0.312 0.019
2008 | -0.120 0.095 -1.263 0.207 -0.307 0.066
2010 | -0.233 0.098 -2.371 0.018 -0.425 -0.040
2012 | -0.068 0.097 -0.704 0.481 -0.258 0.122
2014 | 0.081 0.088 0.925 0.355 -0.091 0.253
2016 | -0.032 0.088 -0.368 0.713 -0.205 0.140
|
age | -0.042 0.011 -3.723 0.000 -0.064 -0.020
|
c.age#c.age | 0.000 0.000 3.709 0.000 0.000 0.001
|
_cons | -0.439 0.242 -1.814 0.070 -0.913 0.035
-----------------------------------------------------------------------------------estimates store m2
logit vhappy i.college c.wages i.married i.parent i.woman i.conserv i.reltrad i.year c.age##c.age,
vce(robust)Iteration 0: Log pseudolikelihood = -5696.026
Iteration 1: Log pseudolikelihood = -5397.02
Iteration 2: Log pseudolikelihood = -5393.5302
Iteration 3: Log pseudolikelihood = -5393.5286
Iteration 4: Log pseudolikelihood = -5393.5286
Logistic regression Number of obs = 9,216
Wald chi2(22) = 568.57
Prob > chi2 = 0.0000
Log pseudolikelihood = -5393.5286 Pseudo R2 = 0.0531
-----------------------------------------------------------------------------------
| Robust
vhappy | Coefficient std. err. z P>|z| [95% conf. interval]
------------------+----------------------------------------------------------------
college |
College Degree | 0.178 0.054 3.273 0.001 0.071 0.285
wages | 0.011 0.002 6.055 0.000 0.007 0.015
|
married |
Married | 0.979 0.052 18.761 0.000 0.877 1.081
|
parent |
Parent | -0.073 0.059 -1.235 0.217 -0.189 0.043
|
woman |
Women | 0.138 0.048 2.844 0.004 0.043 0.232
|
conserv |
Conservative | 0.184 0.050 3.638 0.000 0.085 0.282
|
reltrad |
Mainline Prot | -0.108 0.077 -1.401 0.161 -0.259 0.043
Black Protestant | -0.161 0.101 -1.605 0.108 -0.359 0.036
Catholic | -0.148 0.065 -2.277 0.023 -0.276 -0.021
Jewish | -0.704 0.198 -3.565 0.000 -1.092 -0.317
Other Faith | -0.303 0.111 -2.720 0.007 -0.522 -0.085
Nonaffiliated | -0.283 0.074 -3.842 0.000 -0.428 -0.139
|
year |
2002 | 0.050 0.102 0.493 0.622 -0.150 0.251
2004 | -0.088 0.106 -0.830 0.406 -0.295 0.119
2006 | -0.150 0.084 -1.774 0.076 -0.315 0.016
2008 | -0.123 0.096 -1.283 0.200 -0.310 0.065
2010 | -0.222 0.099 -2.252 0.024 -0.415 -0.029
2012 | -0.055 0.097 -0.571 0.568 -0.246 0.135
2014 | 0.090 0.088 1.023 0.306 -0.083 0.263
2016 | -0.030 0.088 -0.345 0.730 -0.203 0.142
|
age | -0.052 0.011 -4.568 0.000 -0.074 -0.030
|
c.age#c.age | 0.001 0.000 4.393 0.000 0.000 0.001
|
_cons | -0.332 0.243 -1.364 0.172 -0.808 0.145
-----------------------------------------------------------------------------------estimates store m3
logit vhappy i.college c.wages c.occprest i.married i.parent i.woman i.conserv i.reltrad i.year c.
age##c.age, vce(robust)Iteration 0: Log pseudolikelihood = -5696.026
Iteration 1: Log pseudolikelihood = -5390.3082
Iteration 2: Log pseudolikelihood = -5386.6127
Iteration 3: Log pseudolikelihood = -5386.6108
Iteration 4: Log pseudolikelihood = -5386.6108
Logistic regression Number of obs = 9,216
Wald chi2(23) = 575.85
Prob > chi2 = 0.0000
Log pseudolikelihood = -5386.6108 Pseudo R2 = 0.0543
-----------------------------------------------------------------------------------
| Robust
vhappy | Coefficient std. err. z P>|z| [95% conf. interval]
------------------+----------------------------------------------------------------
college |
College Degree | 0.096 0.058 1.640 0.101 -0.019 0.210
wages | 0.010 0.002 5.072 0.000 0.006 0.013
occprest | 0.007 0.002 3.778 0.000 0.004 0.011
|
married |
Married | 0.970 0.052 18.568 0.000 0.867 1.072
|
parent |
Parent | -0.069 0.059 -1.166 0.244 -0.185 0.047
|
woman |
Women | 0.126 0.049 2.587 0.010 0.030 0.221
|
conserv |
Conservative | 0.185 0.051 3.656 0.000 0.086 0.284
|
reltrad |
Mainline Prot | -0.115 0.077 -1.493 0.136 -0.266 0.036
Black Protestant | -0.151 0.101 -1.498 0.134 -0.348 0.047
Catholic | -0.143 0.065 -2.201 0.028 -0.271 -0.016
Jewish | -0.711 0.199 -3.574 0.000 -1.101 -0.321
Other Faith | -0.307 0.112 -2.749 0.006 -0.526 -0.088
Nonaffiliated | -0.282 0.074 -3.819 0.000 -0.426 -0.137
|
year |
2002 | 0.054 0.102 0.526 0.599 -0.146 0.254
2004 | -0.089 0.106 -0.839 0.401 -0.295 0.118
2006 | -0.145 0.084 -1.714 0.087 -0.310 0.021
2008 | -0.118 0.096 -1.233 0.218 -0.305 0.070
2010 | -0.214 0.099 -2.166 0.030 -0.407 -0.020
2012 | -0.054 0.097 -0.551 0.581 -0.244 0.137
2014 | 0.098 0.088 1.115 0.265 -0.075 0.271
2016 | -0.021 0.088 -0.237 0.812 -0.194 0.152
|
age | -0.054 0.011 -4.731 0.000 -0.076 -0.031
|
c.age#c.age | 0.001 0.000 4.571 0.000 0.000 0.001
|
_cons | -0.569 0.251 -2.270 0.023 -1.060 -0.078
-----------------------------------------------------------------------------------estimates store m4. mecompare college, models(m1 m2 m3 m4)Predicting: Pr(vhappy)
Marginal effects across models (N_m1=9216) (N_m2=9216) (N_m3=9216) (N_m4=9216)
| ME # Estimate Robust SE P>|z|
---------------------------------+---------------------------------------
college |
College Deg - No Col Deg |
m1 | 1 0.072 0.010 0.000
m2 | 2 0.060 0.011 0.000
m3 | 3 0.036 0.011 0.001
m4 | 4 0.019 0.012 0.103Use metest to calculate the cross-model tests. Here, whether the effect diminishes in each subsequent model.
metest 1 - 2 | estimate se pvalue
---------------------------------+--------------------------------
college |
m1 - m2 | 0.012 0.004 0.003 metest 2 - 3, add | estimate se pvalue
---------------------------------+--------------------------------
college |
m1 - m2 | 0.012 0.004 0.003
m2 - m3 | 0.024 0.004 0.000 metest 3 - 4, add | estimate se pvalue
---------------------------------+--------------------------------
college |
m1 - m2 | 0.012 0.004 0.003
m2 - m3 | 0.024 0.004 0.000
m3 - m4 | 0.017 0.004 0.000