Adding lincom postestimation results to esttab regression table

[Wouter Wakker]

Dear Statalisters,

I am trying to add some postestimation results from lincom to a regression table using esttab (from SJ). For this I use lincomest (from SSC) which stores the estimation results. What I would want is to have the lincomest results below the coefficients of the regression. I have tried using estadd but haven't managed to get it to work so far. Is this possible? I'm trying to avoid using scalars because then it is impossible to add significance stars.

sysuse auto, clear
eststo: reg price c.mpg##foreign
eststo: lincomest _b[mpg] + _b[1.foreign#c.mpg]
esttab, nogaps nobase cells(b(fmt(3)) se(fmt(3)))


--------------------------------------
                      (1)          (2)
                    price        price
                     b/se         b/se
--------------------------------------
mpg              -329.255            
                   74.985            
1.foreign         -13.587            
                 2634.664            
1.foreign#~g       78.888            
                  112.481            
(1)                           -250.367
                                83.840
_cons           12600.538            
                 1527.888            
--------------------------------------
N                      74           74
--------------------------------------

[Andrew Musau]

It would be difficult to combine the estimates as they pertain to two different models. However,

I'm trying to avoid using scalars because then it is impossible to add significance stars

this is not true. Here, what matters is sequence because you want the lincom results to be linked to your regression estimates, and you can only obtain the lincom results after the regression. This suggests that you need to run the regression at least twice, which is what I do below. With Stata's margins command, you no longer need lincomest from SSC.

sysuse auto, clear
qui reg price c.mpg##foreign
margins, expression(_b[mpg] + _b[1.foreign#c.mpg]) post
mat r= r(table)
eststo: reg price c.mpg##foreign
estadd local mystat "`= cond(r[4,1]<0.01,"`:di %5.3f `=r[1,1]''***", ///
cond(r[4,1]<0.05,"`:di %5.3f `=r[1,1]''**", ///
cond(r[4,1]<0.1,"`:di %5.3f `=r[1,1]''*",  "`:di %5.3f `=r[1,1]''")))'"
esttab, nogaps nobase cells(b(fmt(3)) se(fmt(3))) scalar("mystat My stat")

Res.:

. esttab, nogaps nobase cells(b(fmt(3)) se(fmt(3))) scalar("mystat My stat")

-------------------------
                      (1)
                    price
                     b/se
-------------------------
mpg              -329.255
                   74.985
1.foreign         -13.587
                 2634.664
1.foreign#~g       78.888
                  112.481
_cons           12600.538
                 1527.888
-------------------------
N                      74
My stat       -250.367***
-------------------------

[Wouter Wakker]

Thank you Andrew, I didn't realize you could add significance stars like that. Very helpful.

[Andrew Musau]

See also

https://www.statalist.org/forums/for...dd-and-reghdfe

for a command that does #2.

[Barbora Sedova]

Dear Andrew,

how can I produce a similar table as you show in post 2, however if I have two or mor einteraction terms?
Thank you!

[carmen campagnolo]

Hi Andrew, would you have any suggestions on how to include the SE in parenthesis for the linear combination from post 2? And to move these results to be located above the regression stats (r2 etc)?

Thank you.

[Wouter Wakker]

carmen campagnolo Since this post I've made a package xlincom to assist with these kinds of things, maybe you find it useful.

ssc install xlincom
sysuse auto, clear
reg price c.mpg##foreign
xlincom lin_com = _b[mpg] + _b[1.foreign#c.mpg], repost
esttab, nogaps nobase eqlabels("Main" "Linear combinations", span)

[carmen campagnolo]

Wouter Wakker thanks for the response. Does this package work for Stata 16?

I seem to get an error message when trying your code.

xlincom lin_com = _b[mpg] + _b[1.foreign#c.mpg], repost

lin_com: _b[mpg] + _b[1.foreign#c.mpg] = 0
foreign#co: operator invalid
r(198);

end of do-file

Would you know why this is?

[Wouter Wakker]

This is very strange, I can run the code from #7 fine in Stata 15, 16 and 17.

Could you do set trace on and post the last couple of lines before getting the error when you run the code?

[danishussalam]

Hi Andrew Musau

I am using part of your code to add lincom coefficients with the associated significance stars. After the fourth line of the code below, STATA shows the following:

added macro:
             e(mystat) : "-0.115***"

but the outreg2 doesn't read the e(mystat) value in the addstat section. It gives me an error saying `invalid syntax'. Any idea on how I can fix this? My end goal is to have both coeffient and associated significance stars in the same cell, exactly like it is shown in e(mystat)


reghdfe att treat ebola treat_ebola if year==0, vce(robust) noabsorb
lincomest treat + treat_ebola
mat r= r(table)
matrix list r(table)
estadd local mystat "`= cond(r[4,1]<0.01,"`:di %5.3f `=r[1,1]''***", cond(r[4,1]<0.05,"`:di %5.3f `=r[1,1]''**", cond(r[4,1]<0.1,"`:di %5.3f `=r[1,1]''*", "`:di %5.3f `=r[1,1]''")))'"
outreg2 using table_a3, dec(3) nonotes nocons tex replace addstat(Attrition for Ebola=1, `e(mystat)') keep(treat treat_ebola) label

[Andrew Musau]

The added stars make the contents of the macro string. So you need:

addtext(Attrition for Ebola=1, `e(mystat)')

[danishussalam]

Thanks a bunch! Andrew Musau

[danishussalam]

Hi Andrew Musau there seems to be a problem

The code works fine but it no longer reports the treat_ebola coefficient in the second regression. I am pasting my full code below.


// Reg 1
reghdfe att treat if year==0, vce(robust) noabsorb
outreg2 using table_a3, replace dec(3) adjr2 nonotes tex label

// Reg 2
reghdfe gender treat ebola treat_ebola if year==0, vce(robust) noabsorb
lincomest treat + treat_ebola
mat r= r(table)
estadd local mystat "`= cond(r[4,1]<0.01,"`:di %5.3f `=r[1,1]''***", cond(r[4,1]<0.05,"`:di %5.3f `=r[1,1]''**", cond(r[4,1]<0.1,"`:di %5.3f `=r[1,1]''*", "`:di %5.3f `=r[1,1]''")))'"
outreg2 using table_a3, dec(3) nonotes nocons tex append addtext(Attrition for Ebola=1, `e(mystat)') keep(treat treat_ebola) label

I want something like this (figure below) with starts on "Attrition for Ebola=1"

with the code above what I get is as follows (figure below)

[Andrew Musau]

I cannot follow this. Attach some data example that reproduces the issue.

[danishussalam]

Andrew Musau Here you go:

* Example generated by -dataex-. For more info, type help dataex
clear
input float(year att) double(treat gender) float(ebola treat_ebola)
0 0 0 0 0 0
0 0 1 0 1 1
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 1 1 1
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 0 1 1
0 0 1 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 1 1 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 1 0 0
0 0 1 0 0 0
0 1 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 0 1 1
0 0 0 0 0 0
0 0 0 1 1 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 1 0
0 1 0 1 1 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 1 1
0 0 1 1 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 1 0
0 0 1 0 0 0
0 0 1 0 1 1
0 0 1 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 1 0
0 0 1 0 1 1
0 0 0 0 1 0
0 0 0 0 0 0
0 0 1 0 0 0
0 0 0 1 1 0
0 0 1 0 0 0
0 0 1 0 0 0
0 0 1 0 0 0
0 1 0 0 0 0
0 0 0 0 1 0
0 0 1 0 1 1
0 0 0 0 1 0
0 0 1 0 1 1
0 0 0 0 1 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 1 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 1 0 0
0 0 1 0 1 1
0 0 0 0 0 0
0 0 0 0 1 0
0 0 0 0 1 0
0 0 0 0 1 0
0 0 1 0 0 0
0 0 1 0 1 1
0 0 0 0 0 0
end
label values year year
label def year 0 "2015", modify
label values att att
label def att 0 "Retained", modify
label def att 1 "Attrited", modify
label values treat treat
label def treat 0 "Control", modify
label def treat 1 "Treatment", modify
label values gender gender
label def gender 0 "Female", modify
label def gender 1 "Male", modify