datasummary_balance does not use siunitx

[vincentarelbundock]

Problem is that the final table is a combination of several tables produced by output="data.frame", so the rounding is done in a pre-processing stage, without siunitx.

[acarril]

This is still an issue on modelsummary_1.4.5. Any insights as to how difficult it would be to fix? I could try to tackle this if you accept PRs.

In relation to this, the documentation for datasummary_balance mentions the necessity for the siunitx Latex package and that the default behavior is to enclose numbers in \num{}, which is not happening. Maybe it would be a good idea to remove these mentions from the documentation until this issue is fixed, or at least link to it?

modelsummary/man/datasummary_balance.Rd

Lines 183 to 185 in 2499b55

	By default, LaTeX tables enclose all numeric entries in the \verb{\\num\{\}} command
	from the siunitx package. To prevent this behavior, or to enclose numbers
	in dollar signs (for LaTeX math mode), users can call:

[vincentarelbundock]

Ah yeah, that would be great. I'm sure the solution only takes a few lines of code, but finding the right spot in the logic might (or might not) be tricky.

If you look in the main function, there are calls to datasummary(output="dataframe"). If you dig, you'll find that the numeric formatter disables siuntix wrapping when the output is a dataframe.

This is important, usually, but in this case it bites us.

What we need is to find a way to disable this check, and let datasummary() format with \num{} in the special case where the original calling function is datasummary_balance()

[acarril]

I see, thanks for the details. I'll try to work on it this weekend and see if I can come up with a solution.

[acarril]

What we need is to find a way to disable this check, and let datasummary() format with \num{} in the special case where the original calling function is datasummary_balance()

If this is an approach that you think would work, maybe it can be implemented fairly quickly using something like sys.call() to inspect the call stack. For instance, consider

foo <- function() {
  caller <- sys.call(-1)  # sys.call(-1) returns immediate caller
  return(caller)
}

bar <- function() {
  foo()
}

Here, running foo() would return NULL (i.e. no parent caller), but running bar() returns bar() (the caller):

r$> foo()
NULL

r$> bar()
bar()

I am not familiar with this codebase, but by taking a quick look, I suppose that the "numeric formatter" that you mention is in fmt_factory.R. I'd have to look more closely at how datasummary() is formatting its output. Let me know if I'm missing something!

[vincentarelbundock]

Hmm, maybe this would work. But there's also a package-specific settings_get()/settings_set() mechanism that we can use to pass this kind of information across function calls. So introspecting the call stack might be overkill.

Here's a start of a pull request with a simple change that seems to work. Need to check why some tests are failing and test various cases, but that's a start...

#720

[vincentarelbundock]

Alright, I think I’ve got this working now. It requires the development versions of both `tinytable:

remotes::install_github("vincentarelbundock/tinytable")
remotes::install_github("vincentarelbundock/modelsummary")

This ended up being one of those things that fell into a rather arcane corner of the package, so it was easier to fix myself.

That said, I really appreciate your offer to help! And I really don’t want my taking over the issue to discourage you from making contributions if you run into other issues that you might care about.

I tagged a couple issues as “good first” if you or anyone else want to try something. https://github.com/vincentarelbundock/modelsummary/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22

No pressure, obviously!

library(modelsummary)

dat <- transform(
  mtcars,
  cyl = factor(cyl),
  gear = factor(gear)
)

datasummary_balance(~am, data = dat, stars = TRUE, output = "latex")

\begin{table}
\centering
\begin{tblr}[         %% tabularray outer open
]                     %% tabularray outer close
{                     %% tabularray inner open
colspec={Q[]Q[]Q[]Q[]Q[]Q[]Q[]Q[]},
cell{1}{3}={c=2,}{halign=c,},
cell{1}{5}={c=2,}{halign=c,},
column{1}={halign=l,},
column{2}={halign=l,},
column{3}={halign=r,},
column{4}={halign=r,},
column{5}={halign=r,},
column{6}={halign=r,},
column{7}={halign=r,},
column{8}={halign=r,},
row{1}={halign=c,},
hline{11}={1,2,3,4,5,6,7,8}{solid, 0.05em, black},
}                     %% tabularray inner close
\toprule
&  & 0 &  & 1 &  &  &  \\ \cmidrule[lr]{3-4}\cmidrule[lr]{5-6}
&    & Mean & Std. Dev. & Mean & Std. Dev. & Diff. in Means & Std. Error \\ \midrule %% TinyTableHeader
mpg  &   & \num{17.1}  & \num{3.8}   & \num{24.4}  & \num{6.2}  & \num{7.2}**     & \num{1.9}  \\
disp &   & \num{290.4} & \num{110.2} & \num{143.5} & \num{87.2} & \num{-146.8}*** & \num{35.0} \\
hp   &   & \num{160.3} & \num{53.9}  & \num{126.8} & \num{84.1} & \num{-33.4}     & \num{26.4} \\
drat &   & \num{3.3}   & \num{0.4}   & \num{4.0}   & \num{0.4}  & \num{0.8}***    & \num{0.1}  \\
wt   &   & \num{3.8}   & \num{0.8}   & \num{2.4}   & \num{0.6}  & \num{-1.4}***   & \num{0.2}  \\
qsec &   & \num{18.2}  & \num{1.8}   & \num{17.4}  & \num{1.8}  & \num{-0.8}      & \num{0.6}  \\
vs   &   & \num{0.4}   & \num{0.5}   & \num{0.5}   & \num{0.5}  & \num{0.2}       & \num{0.2}  \\
carb &   & \num{2.7}   & \num{1.1}   & \num{2.9}   & \num{2.2}  & \num{0.2}       & \num{0.7}  \\
&   & N            & Pct.         & N            & Pct.        &                  &             \\
cyl  & 4 & \num{3}     & \num{15.8}  & \num{8}     & \num{61.5} &                  &             \\
& 6 & \num{4}     & \num{21.1}  & \num{3}     & \num{23.1} &                  &             \\
& 8 & \num{12}    & \num{63.2}  & \num{2}     & \num{15.4} &                  &             \\
gear & 3 & \num{15}    & \num{78.9}  & \num{0}     & \num{0.0}  &                  &             \\
& 4 & \num{4}     & \num{21.1}  & \num{8}     & \num{61.5} &                  &             \\
& 5 & \num{0}     & \num{0.0}   & \num{5}     & \num{38.5} &                  &             \\
\bottomrule
\end{tblr}
\end{table} 

[acarril]

Awesome! Thank you for the fix.

That said, I really appreciate your offer to help! And I really don’t want my taking over the issue to discourage you from making contributions if you run into other issues that you might care about.

Not discouraged at all, I appreciate it that you took the time do it. I'll keep my eyes peeled!