In mkvasnicka/stargate: Simple Formatted Regression Tables

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

stargate

The goal of stargate package is to simplify production of formatted tables that hold regression analysis results from several models side-by-side.

For many years, we have used stargazer package for this purpose. stargazer is great and every applied econometrician has to be thankful to Marek Hlaváč for kindly providing it. However, this package is implemented as one monolithic function, which has two drawbacks:

It is very difficult for an user to implement a method for an estimated object not covered in stargazer.
It has too many complex parameters. A set of lightweight manipulation function joined with pipes (in the tidyverse style) seems preferable.
It has its own output filters to LaTeX, HTML, etc., and hence it is impossible to customize output for your own liking.

This package is an attempt to rewrite stargazer remove these obstacles. To pay homage to the original package, it is named similarly: stargate.

Every estimated object class should have its own method to convert the estimate into a standard format. The broom is used now but dedicated methods would be provided later to customize the process, e.g.\ to include robust SEs.
The workflow must be based on simples function that send their outputs between themselves through pipes.
The final output must be a tibble (convertible to data.frame, etc.), and hence everyone can create the final table with any tool that suits him or her, e.g.\ kbl() from kableExtra.

However, this package is just in a rudimentary state---it is more a proof of the concept and a playground than a real thing. It is here to raise some feedback, comments, and wishes. If used, it must be used with extreme care:

It can handle only the estimates handled by broom at the present time.
Anything can change at any time.
Anything can fail you expectations at any time.

You have been warned. Enjoy!

Installation

At the present time, there is only the development version. It is available from GitHub with:

# install.packages("devtools")
devtools::install_github("mkvasnicka/stargate")

Example

This is a basic example which shows you how to use stargate. Load the necessary packages first:

library(stargate)
library(tibble)
library(kableExtra)

Estimates some models (simulated here):

n <- 1e3
df <- tibble(
    x1 = rnorm(n),
    x2 = rnorm(n),
    x3 = rnorm(n),
    e  = rnorm(n),
    y  = x1 - 2 * x2 + 3 * x3 + e
)

m1 <- lm(y ~ x1 + x2, df)
m2 <- lm(y ~ x1 + x2 + x3, df)
m3 <- lm(y ~ x1 + x2 * x3, df)

Convert the estimates to their stargate representation: one model with sg_model(), many models with sg_table():

sm1 <- sg_model(m1)
stab <- sg_table(sm1, m2, m3)

You can also convert at least one estimate to a stargate model or table and then add other using the plus (+) sign:

stab <- sm1 + m2 + m3

Transform the model table to your liking:

stab <- stab |>
    sg_remove(coefs = "Interc", regex = TRUE) |> 
    sg_remove(stats = c("r.squared", "logLik", "nobs"), keep = TRUE)

Rename coefficients and statistics as you like:

stab <- stab |> 
    sg_rename(coefs = c("a" = "x", " x " = ":"), regex = TRUE) |> 
    sg_rename(stats = c("R2" = "r.squared", "log Lik." = "logLik"))

Format the table:

sftab <- stab |> 
    sg_format()

Present:

sftab <- sftab |> 
    sg_present()

Print the table:

kbl(sftab, format = "pipe") |> 
    kable_classic()

Or do all that at once:

sg_table(sm1, m2, m3) |> 
    sg_remove(coefs = "Interc", regex = TRUE) |> 
    sg_remove(stats = c("r.squared", "logLik", "nobs"), keep = TRUE) |> 
    sg_rename(coefs = c("a" = "x", " x " = ":"), regex = TRUE) |> 
    sg_rename(stats = c("R2" = "r.squared", "log Lik." = "logLik",
                        "num. of. obs." = "nobs")) |> 
    sg_format(style = c("{estimate}", "({std.error})"),
              format = list(nsmall = 3)) |> 
    sg_present() |> 
    kbl(format = "pipe") |> 
    kable_classic()

What next

Generalize the sg_model class to be able to encompass all useful information that are stored in model estimates produced by most packages.
Implement the stuff. :-)
Add protective code (checking inputs, etc.).
Implement non-numeric coefficients (as Yes/No for included fixed effects)---and the way they are formatted.
Possibly add classes of coefficients (constant, basic, interaction, special, note, etc.) to simplify their ordering.
Provide methods for estimated objects.
Write a vignette how to write a new method.
Provide some auto-setup for the function to simplify the process.
Tune the output to LaTeX, HTML, Markdown, and possibly Word/LibreOffice.
Implement print() methods for all stargate objects.