Add headers to groups of variables in output from estout?

[Michael Anbar]

I have several specifications that -estout- groups into a coefficient table. I'm using -estout- version 3.20 from the SSC and the March 3, 2016 update of Stata 14.1 MP. So far, I have code like this:

sysuse auto, clear

gen mpg2 = mpg^2
gen mpg_log = log(mpg)

regress price mpg* headroom
est store est1

regress price mpg* headroom length
est store est2

estout est?, drop(headroom)

which displays

--------------------------------------
                     est1         est2
                        b            b
--------------------------------------
mpg              4073.189     4132.779
mpg2            -33.14309    -33.83749
mpg_log         -61992.16    -62350.48
length                        3.698642
_cons            124383.9     123889.8
--------------------------------------

I'd like to have output that looks like this:

--------------------------------------
                     est1         est2
                        b            b
--------------------------------------
Mileage variables
mpg                      4073.189     4132.779
mpg2                    -33.14309    -33.83749
mpg_log                 -61992.16    -62350.48

length                        3.698642

_cons                    124383.9     123889.8
--------------------------------------

This is a rough approximation because I created it manually, but basically, I'd like to group certain variables under a header, within the first column, and be able to separate groups of variables with empty lines in the table. I'm sure there's a way to do this manually, by creating empty observations, but it's not clear to me from the -estout- documentation if there is a way to use -estout- for this.

[Nils Enevoldsen]

http://repec.org/bocode/e/estout/hlp_estout.html#refcat

[Michael Anbar]

http://repec.org/bocode/e/estout/hlp_estout.html#refcat

Nils Enevoldsen Does this support adding line breaks between categories? This code

estout est?, drop(headroom) refcat(mpg "Mileage" length "Other", nolabel)

displays

--------------------------------------
                     est1         est2
                        b            b
--------------------------------------
Mileage                              
mpg              4073.189     4132.779
mpg2            -33.14309    -33.83749
mpg_log         -61992.16    -62350.48
Other                                
length                        3.698642
_cons            124383.9     123889.8
--------------------------------------

which is much closer to what I need, but it's not perfect. I thought about adding reference category labels below variables, e.g. -mpg_log- in this case, but the -refcat- option doesn't support placing some labels above variables and other labels below. Compound quotes, i.e.

estout est?, drop(headroom) refcat(mpg `""" "Mileage""' length "Other", nolabel)

and manual line breaks, i.e. prefixing the string with "\n" or "{break}", don't work either.

My goal is a table like this (for example):

--------------------------------------
                     est1         est2
                        b            b
--------------------------------------
Mileage                          
mpg              4073.189     4132.779
mpg2            -33.14309    -33.83749
mpg_log         -61992.16    -62350.48

Other                          
length                        3.698642
_cons            124383.9     123889.8
--------------------------------------

[Nils Enevoldsen]

Two suggestions, assuming you're exporting to LaTeX:

1. Put something like \rule{0pt}{4ex} before the category name to add some vertical space
2. Create a new "empty" regressor with order(mpg mpg2 mpg_log "\ ")

If you're not using LaTeX, let me know, and we can come up with some other language-specific hack.

[Michael Anbar]

Nils Enevoldsen Unfortunately, I'm not exporting to LaTeX; I'm exporting to text and tab-delimited files (specifically, the -fixed- style and the -tab- styles) and displaying the table in Stata (so the default SMCL style). Adding an extra regressor doesn't appear to work for these styles.

[Nils Enevoldsen]

For SMCL, I think adding a regressor named "{break}" will work. Can you try that? I'm still not sure what to do about tab-delimited, though.

[Michael Anbar]

For SMCL, I think adding a regressor named "{break}" will work. Can you try that? I'm still not sure what to do about tab-delimited, though.

That works for SMCL, but the only solution I can think of for tab-delimited or text files is to load the file into Stata as-is (read the text file into a dataset) and add observations, i.e. table rows, and blank out the lines. That's a clumsy solution at best, though.

[Alessio Gaggero]

Two suggestions, assuming you're exporting to LaTeX:

1. Put something like \rule{0pt}{4ex} before the category name to add some vertical space
2. Create a new "empty" regressor with order(mpg mpg2 mpg_log "\ ")

If you're not using LaTeX, let me know, and we can come up with some other language-specific hack.

Dear Nils,

I was wondering if you can expand on this a little bit. How exactly do you create an "empty" regressor?

thanks or the help
Alessio

[Nils Enevoldsen]

The "empty" regressor in order(mpg mpg2 mpg_log "\ ") is the "\ ". If you have order(mpg mpg2 mpg_log foo) then esttab will show the regressors "mpg", "mpg2", "mpg_log", and "foo". "Foo" won't have any coefficients listed, because they don't exist. If you have order(mpg mpg2 mpg_log "\ ") then esttab will show the regressors "mpg", "mpg2", "mpg_log", and " ".