Regression tables

modelsummary makes tables from a combined object and from the marginaleffects results computed with it. Add shape = term ~ group to put each model in its own column.

The models are the two logits of union membership from Getting started:

library(suest)
library(marginaleffects)
library(modelsummary)
library(haven)

nlsw <- zap_labels(read_dta("https://tdmize.github.io/data/data/nlsw88.dta"))
nlsw <- nlsw[!is.na(nlsw$union), ]
nlsw[c("collgrad", "south")] <- lapply(nlsw[c("collgrad", "south")], factor)

base <- glm(union ~ collgrad + age + south, family = binomial("logit"),
            data = nlsw)
adjusted <- glm(union ~ collgrad + age + south + wage + ttl_exp,
                family = binomial("logit"), data = nlsw)
combined <- suest(base, adjusted, model_names = c("Base", "Adjusted"))

Coefficients

The standard errors are each model’s robust standard errors from the joint covariance matrix.

modelsummary(combined, shape = term ~ group)

	(1)
	Base	Adjusted
(Intercept)	-1.340	-1.761
	(0.704)	(0.719)
collgrad1	0.507	0.311
	(0.119)	(0.132)
age	0.009	0.008
	(0.018)	(0.018)
south1	-0.747	-0.677
	(0.116)	(0.118)
wage		0.054
		(0.014)
ttl_exp		0.006
		(0.013)
Num.Obs.	1878

Marginal effects and their differences

effects <- avg_slopes(combined, variables = c("collgrad", "age", "south"))
differences <- hypotheses(effects, hypothesis = difference ~ revpairwise | term)
differences$group <- "Base - Adjusted"

modelsummary(
  list("Average marginal effects" = effects, "Difference" = differences),
  shape = term ~ group
)

	Average marginal effects		Difference
	Base	Adjusted	Base - Adjusted
age	0.002	0.001	0.000
	(0.003)	(0.003)	(0.001)
collgrad	0.097	0.057	0.039
	(0.024)	(0.025)	(0.009)
south	-0.131	-0.117	-0.013
	(0.019)	(0.020)	(0.004)
Num.Obs.	1878		1878