customize numbers in modelsummary estimate

Question

I can use the following code

library(fixest)
library(modelsummary)

reg1 <- feols(mpg ~ cyl , data = mtcars)
reg2 <- feols(mpg ~ cyl + wt , data = mtcars)



modelsummary(
  list(reg1, reg2),
  fmt = NULL,
  estimate  = c("{round(estimate, 3)} [my number]"),
  statistic = "{round(std.error, 3)}",
  coef_omit = "Intercept",
  gof_omit = 'RMSE|AIC|BIC|R2|Std.Errors')

to display numbers in square brackets next to my estimates. However I want different numbers for each of the coefficient in each of the regressions. Can I do this? the estimate argument does not seem to accept vector or list inputs.

Answer 1

There is no built-in way to round numbers differently on a per-model basis. I don’t know any regression table package that can do that, but perhaps I’m missing something. That said, it is reasonably easy to do it semi-manually:

1. Export each model to a modelsummary_list object.
2. Round the numbers as you want.
3. Feed the modelsummary_list objects back to the modelsummary() function.

library(modelsummary)
mod1 <- lm(mpg ~ cyl + disp, mtcars)
mod2 <- lm(mpg ~ cyl + disp + hp, mtcars)

extract_and_round <- function(model, significant_digits = 2) {
    out <- modelsummary(model, output = "modelsummary_list")
    out$tidy$estimate <- format(out$tidy$estimate, digits = significant_digits)
    out$tidy$std.error <- format(out$tidy$std.error, digits = significant_digits)
    return(out)
}

models <- list(
    extract_and_round(mod1, 2),
    extract_and_round(mod2, 5)
)

# Feed the modelsummary_list objects back to modelsummary()
modelsummary(models, output = "markdown")

	(1)	(2)
(Intercept)	34.661	34.184919
	(2.55)	(2.590778)
cyl	-1.587	-1.227420
	(0.71)	(0.797276)
disp	-0.021	-0.018838
	(0.01)	(0.010404)
hp		-0.014679
		(0.014651)
Num.Obs.	32	32
R2	0.760	0.768
R2 Adj.	0.743	0.743
AIC	167.1	168.0
BIC	173.0	175.3
Log.Lik.	-79.573	-79.009
F	45.808	30.877
RMSE	2.91	2.86