Removing certain parts of modelsummary in R at specific statistics - r

Question

I am using gamm4:gamm4 to model longitudinal change.
I am trying to use the modelsummary::modelsummary function to create an output table of the following results:
I would like to add t-values and std.error to the output of the fixed effects, and remove the empty tags values from the random effects
model_lmer <- gamm4(Y ~ Tract + s(Age, by = Tract, k = 10) + Sex,
data = (DF1),
random = ~ (0 + Tract | ID))
modelsummary(model_lmer$mer,
statistic = c("s.e. = {std.error}",
"t = {statistic}"))
But I am struggling to write the correct syntax to remove the "t" and "s.e." from the random effects output.

Answer 1

This is kind of tricky, actually. The issue is that modelsummary()
automatically drops empty rows when they are filled with NA or an
empty string "". However, since glue strings can include arbitrary
text, it is hard to think of a general way to figure out if the row is
empty or not, because modelsummary() cannot know ex ante what
constitutes an empty string.
If you have an idea on how this check could be implemented, please report it
on Github:
https://github.com/vincentarelbundock/modelsummary
In the meantime, you could use the powerful tidy_custom.CLASSNAME
mechanism
to customize the statistic and p.value statistics directly instead
of through a glue string:
library(gamm4)
library(modelsummary)
# simulate
x <- runif(100)
fac <- sample(1:20,100,replace=TRUE)
eta <- x^2*3 + fac/20; fac <- as.factor(fac)
y <- rpois(100,exp(eta))
# fit
mod <- gamm4(y~s(x),family=poisson,random=~(1|fac))
# customize
tidy_custom.glmerMod <- function(x) {
out <- parameters::parameters(x)
out <- insight::standardize_names(out, style = "broom")
out$statistic <- sprintf("t = %.3f", out$statistic)
out$p.value <- sprintf("p = %.3f", out$p.value)
out
}
# summarize
modelsummary(mod$mer,
statistic = c("{statistic}", "{p.value}"))
Model 1
X(Intercept)
1.550
t = 17.647
p = 0.000
Xs(x)Fx1
0.855
t = 4.445
p = 0.000
Num.Obs.
100
RMSE
2.49
Note that I used simple glue strings in statistic = "{p.value}", otherwise they would be wrapped up in parentheses, as is default for standard errors.