Assigning different coefficient names and goodness of fit statistics to different panels in modelsummary

Question

I am trying to make a table using the modelsummary package in R (thank you to the creator) which has different goodness of fit statistics and names for different panels. I have been unable to achieve this so far. I am working in Quarto.

Here is an example (never mind the validity of the regression). I would like to have the adjusted R squared in the first panel, but not the second. I would also like the coefficients to be renamed as in the coefficient mapping, but this does not seem to be happening.

Very grateful for your help.

library(modelsummary)
gm<-tibble::tribble(~raw,~clean,~fmt,
                    "nobs","N",0,
                    "adj.r.squared","Adj. $R^2$",2)

cm <- c("qsec" = "OLS",
        "fit_qsec" = "TSLS")


panels <- list(
  "OLS" = list(
    "(I)" = feols(mpg ~ 1 + qsec, data = mtcars),
    "(II)" = feols(hp ~ 1 + qsec, data = mtcars)
  ),
  "TSLS" = list(
    "(I)" = feols(mpg ~ 1| qsec~drat,data = mtcars),
    "(II)" = feols(hp ~ 1| qsec~drat,data = mtcars)
  )
)

modelsummary(
  panels,
  shape = "rbind",
  cof_map = cm,
  gof_map = gm)

Answer 1

Thanks for the question.

First, there’s a typo in your example: cof_map.

Second, there’s not a built-in way to use different arguments like gof_map in different panels. However, maybe you can hack your way there by using the glance_custom() extension mechanism. For this, you’ll need the development version from github:

remotes::install_github("vincentarelbundock/modelsummary")

Restart R entirely, then define a glance_custom.fixest() which includes a conditional check to delete the adj.r.squared value if it meets some condition. Here, we check if this is an instrumental variables model (bottom panel):

library(fixest)
library(modelsummary)

cm <- c(
    "qsec" = "OLS",
    "fit_qsec" = "TSLS")

panels <- list(
    "OLS" = list(
        "(I)" = feols(mpg ~ 1 + qsec, data = mtcars),
        "(II)" = feols(hp ~ 1 + qsec, data = mtcars)
    ),
    "TSLS" = list(
        "(I)" = feols(mpg ~ 1 | qsec ~ drat, data = mtcars),
        "(II)" = feols(hp ~ 1 | qsec ~ drat, data = mtcars)
    )
)

glance_custom.fixest <- function(x, ...) {
    if (isTRUE(x$iv)) {
        out <- data.frame("adj.r.squared" = NA)
    } else {
        out <- NULL
    }
    return(out)
}

modelsummary(
    panels,
    shape = "rbind",
    coef_map = cm,
    gof_map = c("nobs", "adj.r.squared"))

	(I)	(II)
OLS	1.412	-27.174
	(0.559)	(4.946)
Num.Obs.	32	32
R2 Adj.	0.148	0.485
TSLS	25.190	-188.788
	(47.992)	(328.033)
Num.Obs.	32	32