RegressionTables.jl

This package provides publication-quality regression tables for use with FixedEffectModels.jl and GLM.jl, as well as any package that implements the RegressionModel abstraction.

In its objective it is similar to (and heavily inspired by) the Stata command esttab and the R package stargazer.

Table of Contents

• Installation
• A brief demonstration
• Function Reference
• Frequently Asked Questions

Installation

To install the package, type in the Julia command prompt

] add RegressionTables

A brief demonstration

using RegressionTables, DataFrames, FixedEffectModels, RDatasets, CategoricalArrays

df = dataset("datasets", "iris")
df[!,:SpeciesDummy] = categorical(df[!,:Species])

rr1 = reg(df, @formula(SepalLength ~ SepalWidth + fe(SpeciesDummy)))
rr2 = reg(df, @formula(SepalLength ~ SepalWidth + PetalLength + fe(SpeciesDummy)))
rr3 = reg(df, @formula(SepalLength ~ SepalWidth + PetalLength + PetalWidth + fe(SpeciesDummy)))
rr4 = reg(df, @formula(SepalWidth ~ SepalLength + PetalLength + PetalWidth + fe(SpeciesDummy)))

regtable(rr1,rr2,rr3,rr4; renderSettings = asciiOutput())

yields

----------------------------------------------------------
                         SepalLength            SepalWidth
               ------------------------------   ----------
                    (1)        (2)        (3)          (4)
----------------------------------------------------------
SepalWidth     0.804***   0.432***   0.496***             
                (0.106)    (0.081)    (0.086)             
PetalLength               0.776***   0.829***      -0.188*
                           (0.064)    (0.069)      (0.083)
PetalWidth                            -0.315*     0.626***
                                      (0.151)      (0.123)
SepalLength                                       0.378***
                                                   (0.066)
----------------------------------------------------------
SpeciesDummy        Yes        Yes        Yes          Yes
----------------------------------------------------------
Estimator           OLS        OLS        OLS          OLS
----------------------------------------------------------
N                   150        150        150          150
R2                0.726      0.863      0.867        0.635
----------------------------------------------------------

LaTeX output can be generated by using

regtable(rr1,rr2,rr3,rr4; renderSettings = latexOutput())

which yields

\begin{tabular}{lrrrr}
\toprule
             & \multicolumn{3}{c}{SepalLength} & \multicolumn{1}{c}{SepalWidth} \\
\cmidrule(lr){2-4} \cmidrule(lr){5-5}
             &      (1) &      (2) &       (3) &                            (4) \\
\midrule
SepalWidth   & 0.804*** & 0.432*** &  0.496*** &                                \\
             &  (0.106) &  (0.081) &   (0.086) &                                \\
PetalLength  &          & 0.776*** &  0.829*** &                        -0.188* \\
             &          &  (0.064) &   (0.069) &                        (0.083) \\
PetalWidth   &          &          &   -0.315* &                       0.626*** \\
             &          &          &   (0.151) &                        (0.123) \\
SepalLength  &          &          &           &                       0.378*** \\
             &          &          &           &                        (0.066) \\
\midrule
SpeciesDummy &      Yes &      Yes &       Yes &                            Yes \\
\midrule
Estimator    &      OLS &      OLS &       OLS &                            OLS \\
\midrule
$N$          &      150 &      150 &       150 &                            150 \\
$R^2$        &    0.726 &    0.863 &     0.867 &                          0.635 \\
\bottomrule
\end{tabular}

Similarly, HTML tables can be created with htmlOutput().

Send the output to a text file by passing the destination file string to the asciiOutput(), latexOutput(), or htmlOutput() functions:

regtable(rr1,rr2,rr3,rr4; renderSettings = latexOutput("myoutputfile.tex"))

then use \input in LaTeX to include that file in your code. Be sure to use the booktabs package:

\documentclass{article}
\usepackage{booktabs}

\begin{document}

\begin{table}
\label{tab:mytable}
\input{myoutputfile}
\end{table}

\end{document}

regtable() can also print TableRegressionModel's from GLM.jl (and output from other packages that produce TableRegressionModel's):

using GLM

dobson = DataFrame(Counts = [18.,17,15,20,10,20,25,13,12],
    Outcome = categorical(repeat(["A", "B", "C"], outer = 3)),
    Treatment = categorical(repeat(["a","b", "c"], inner = 3)))
lm1 = fit(LinearModel, @formula(SepalLength ~ SepalWidth), df)
gm1 = fit(GeneralizedLinearModel, @formula(Counts ~ 1 + Outcome + Treatment), dobson,
                  Poisson())

regtable(rr1,lm1,gm1; renderSettings = asciiOutput())

yields

---------------------------------------------
                   SepalLength        Counts
               -------------------   --------
                    (1)        (2)        (3)
---------------------------------------------
(Intercept)    6.526***   6.526***   3.045***
                (0.479)    (0.479)    (0.171)
SepalWidth       -0.223     -0.223           
                (0.155)    (0.155)           
Outcome: B                             -0.454
                                      (0.202)
Outcome: C                             -0.293
                                      (0.193)
Treatment: b                            0.000
                                      (0.200)
Treatment: c                            0.000
                                      (0.200)
---------------------------------------------
Estimator           OLS        OLS         NL
---------------------------------------------
N                   150        150          9
R2                0.014      0.014           
---------------------------------------------

Printing of StatsBase.RegressionModels is experimental; please file as issue if you encounter problems printing them.

Function Reference

Function Arguments

• rr::Union{FixedEffectModel,DataFrames.TableRegressionModel}... are the FixedEffectModels from FixedEffectModels.jl (or TableRegressionModels from GLM.jl) that should be printed. Only required argument.

• regressors is a Vector of regressor names (Strings) that should be shown, in that order. Defaults to an empty vector, in which case all regressors will be shown.

• fixedeffects is a Vector of FE names (Strings) that should be shown, in that order. Defaults to an empty vector, in which case all FE's will be shown. Note that strings need to match the displayed label exactly, otherwise they will not be shown.

• align is a Symbol from the set [:l,:c,:r] indicating the alignment of results columns (default :r right-aligned). Currently affects only latex and ASCII output.

• labels is a Dict that contains displayed labels for variables (strings) and other text in the table. If no label for a variable is found, it default to variable names. See documentation for special values.

• estimformat is a String that describes the format of the estimate. Defaults to "%0.3f".

• estim_decoration is a Function that takes the formatted string and the p-value, and applies decorations (such as the beloved stars). Defaults to (* p<0.05, ** p<0.01, *** p<0.001).

• statisticformat is a String that describes the format of the number below the estimate (se/t). Defaults to "%0.4f".

• below_statistic is a Symbol that describes a statistic that should be shown below each point estimate. Recognized values are :blank, :se, :tstat, and :none. :none suppresses the line. Defaults to :se.

• below_decoration is a Function that takes the formatted statistic string, and applies a decorations. Defaults to round parentheses.

• regression_statistics is a Vector of Symbols that describe statistics to be shown at the bottom of the table. Recognized symbols are :nobs, :r2, :r2_a, :r2_within, :f, :p, :f_kp, :p_kp, and :dof. Defaults to [:nobs, :r2].

• custom_statistics is a NamedTuple that takes user specified statistics to be shown just above regression_statistics. By default each statistic will be labelled by its key (e.g. __LABEL_STATISTIC_mystat__ for the statistic mystat). Defaults to missing. See test/RegressionTables.jl for an example of how to use this.

• number_regressions is a Bool that governs whether regressions should be numbered. Defaults to true.

• number_regressions_decoration is a Function that governs the decorations to the regression numbers. Defaults to s -> "($s)".

• groups is a Vector of labels used to group regressions. This can be useful if results are shown for different data sets or sample restrictions. Defaults to [].

• print_fe_section is a Bool that governs whether a section on fixed effects should be shown. Defaults to true.

• print_estimator_section is a Bool that governs whether to print a section on which estimator (OLS/IV) is used. Defaults to true.

• print_result is a Bool that governs whether the table should be printed to stdout. Defaults to true.

• standardize_coef is a Bool that governs whether the table should show standardized coefficients. Note that this only works with TableRegressionModels, and that only coefficient estimates and the below_statistic are being standardized (i.e. the R^2 etc still pertain to the non-standardized regression).

• out_buffer is an IOBuffer that the output gets sent to (unless an output file is specified, in which case the output is only sent to the file).

• renderSettings::RenderSettings is a RenderSettings composite type that governs how the table should be rendered. Standard supported types are ASCII (via asciiOutput(outfile::String)) and LaTeX (via latexOutput(outfile::String)). If no argument to these two functions are given, the output is sent to STDOUT. Defaults to ASCII with STDOUT.

• transform_labels is a function or a Dict that is used to transform labels. Defaults to identity.

  Some common use cases can be achieved by passing a Symbol instead: :latex, :ampersand, :underscore, :underscore2space. For illustration, here are the three ways to escape forbidden LaTeX characters.

  # Option 1
  regtable(rr; renderSettings = latexOutput(), transform_labels = :latex)
  # Option 2
  repl_dict = Dict("&" => "\\&", "%" => "\\%", "\$" => "\\\$", "#" => "\\#", "_" => "\\_", "{" => "\\{", "}" => "\\}")
  regtable(rr; renderSettings = latexOutput(), transform_labels = repl_dict)
  # Option 3
  function transform(s, repl_dict=repl_dict)
      for (old, new) in repl_dict
          s = replace.(s, Ref(old => new))
      end
      s
  end
  regtable(rr; renderSettings = latexOutput(), transform_labels = transform)
  

Label Codes

The following is the exhaustive list of strings that govern the output of labels. Use e.g.

labels = Dict("__LABEL_STATISTIC_N__" => "Number of observations")

to change the label for the row showing the number of observations in each regression.

• __LABEL_ESTIMATOR__ (default: "Estimator")

• __LABEL_ESTIMATOR_OLS__ (default: "OLS")

• __LABEL_ESTIMATOR_IV__ (default: "IV")

• __LABEL_ESTIMATOR_NL__ (default: "NL")

• __LABEL_FE_YES__ (default: "Yes")

• __LABEL_FE_NO__ (default: "")

• __LABEL_STATISTIC_N__ (default: "N" in asciiOutput())

• __LABEL_STATISTIC_R2__ (default: "R2" in asciiOutput())

• __LABEL_STATISTIC_R2_A__ (default: "Adjusted R2" in asciiOutput())

• __LABEL_STATISTIC_R2_WITHIN__ (default: "Within-R2" in asciiOutput())

• __LABEL_STATISTIC_F__ (default: "F" in asciiOutput())

• __LABEL_STATISTIC_P__ (default: "F-test p value" in asciiOutput())

• __LABEL_STATISTIC_F_KP__ (default: "First-stage F statistic" in asciiOutput())

• __LABEL_STATISTIC_P_KP__ (default: "First-stage p value" in asciiOutput())

• __LABEL_STATISTIC_DOF__ (default: "Degrees of Freedom" in asciiOutput())

Frequently Asked Questions

What's the best way to render regression tables in Pluto.jl?

Use renderSettings = htmlOutput() and print_result = false, and print the resulting String as text/html. This page shows an example.