Supported Models

Adjusted predictions and marginal means

Under the hood, marginaleffects’s predictions and marginalmeans functions use the insight package to retrieve adjusted predictions from a wide variety of models. Currently, insight supports very many model types, and most should work out-of-the-box with the predictions function. If you run into problems, do not hesitate to report them on Github or via email.

Marginal effects and contrasts

This table shows the list of 69 model types for which the marginaleffects() function can compute slopes and contrasts. There are three main alternative software packages to compute such slopes (1) Stata’s margins command, (2) R’s margins::margins function, and (3) R’s emmeans::emtrends function. The test suite hosted on Github compares the numerical equivalence of results produced by marginaleffects to those produced by all 3 alternative software packages:

I am very eager to add support for new models. Feel free to file a request or submit code on Github.

Numerical equivalence
Supported by marginaleffects
Stata
margins
emtrends
Package Function dY/dX SE dY/dX SE dY/dX SE
stats lm
glm
loess U U
AER ivreg U U
tobit U U
afex afex_aov U U
aod betabin U U U U
betareg betareg
bife bife U U U U
biglm biglm U U U U
bigglm U U U U
brglm2 bracl U U U U
brglmFit
brnb U U
brmultinom U U U U
brms brm U U
crch crch U U U U
hxlr U U U U
estimatr lm_lin
lm_robust U
iv_robust U U U U
fixest feols U U U U
feglm U U U U
fenegbin U U U U
fepois U U U U
gam gam U U
gamlss gamlss U U
geepack geeglm U U
glmmTMB glmmTMB U U
glmx glmx U U U
ivreg ivreg U U
lme4 lmer
glmer
glmer.nb
lmerTest lmer
MASS glmmPQL U U
glm.nb
polr
rlm
mclogit mblogit U U U U
mclogit U U U U
mgcv gam U U
bam U U
mhurdle mhurdle U U
mlogit mlogit U U U U
nlme gls U U
nnet multinom U U U U
ordbetareg ordbetareg U U
ordinal clm U U U U
plm plm U U
pscl hurdle U
zeroinfl U
quantreg rq U U
rms lrm U U
robust lmRob U U U U
robustbase glmrob U U
lmrob U U
robustlmm rlmer U U
rstanarm stan_glm U
sampleSelection selection U U U U
heckit U U U U
scam scam U U U U
speedglm speedglm U U
speedlm U U
survey svyglm
survival clogit
coxph U U
tobit1 tobit1 U U
truncreg truncreg U U