Keeping only selected variables in -esttab- output.

[Sonnen Blume]

I am running the following analysis in Stata 14.

quietly mlogit   bmi   i.age  i.smoke   i.beef   i.chicken   if sex=1
 eststo a
 

quietly mlogit     bmi  i.age    i.smoke   i.beef   i.chicken   if sex=2
eststo b


quietly esttab a b, ci  ar2  eform

esttab using BMI.rtf, ci r2  eform

Is it possible to keep only selected variables in esttab, as I want the output only for i.age and i.smoke.

[Weiwen Ng]

esttab a b, ci ar2 eform keep(age smoke)
esttab a b, ci ar2 eform drop(beef chicken)

Both are equivalent in this example.

[Sonnen Blume]

esttab a b, ci ar2 eform keep(age smoke)
esttab a b, ci ar2 eform drop(beef chicken)

Both are equivalent in this example.

Thanks a lot Weiwen. The drop function is working for me. Just installed dataex.

[Magnus Schauf]

Dear members,
I have a similar problem, but for a categorical variable Stata tells me that the "coefficient 'categorical variable' not found".
I used the following code with the -estout- program by Ben Jann:

global xlist1 sales mtb i.industry i.year
eststo clear
foreach var of varlist $ylist {
 eststo: xtreg `var' $xlist1
}
esttab using almeidare.rtf, ci eform drop(industry year) ///
 label scalars(r2_o) onecell nogap modelwidth(10) replace

where industry and year both appear in the global macro xlist1 as categorical variables. This however gives me the error message mentioned above, also if I specify the

 drop(i.industry i.year)

in the -esttab- command

How can I drop the two categorical variables in the export/output?

[Joro Kolev]

I think the problem is that you re not referring to the estimated coefficient/variable with the name that Stata has assigned to it. I.e., Stata does not accept the syntax -drop(i.industry)-.

. set obs 10
number of observations (_N) was 0, now 10

. gen dummy = 1

. replace dummy =2 in 5/l
(6 real changes made)

. gen y = rnormal()

. gen x = rnormal()

. reg y x i.dummy

      Source |       SS           df       MS      Number of obs   =        10
-------------+----------------------------------   F(2, 7)         =      0.60
       Model |  .362903293         2  .181451647   Prob > F        =    0.5768
    Residual |  2.13151275         7  .304501822   R-squared       =    0.1455
-------------+----------------------------------   Adj R-squared   =   -0.0987
       Total |  2.49441605         9  .277157339   Root MSE        =    .55182

------------------------------------------------------------------------------
           y |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
           x |   .0331587   .1742146     0.19   0.854    -.3787933    .4451106
     2.dummy |   .3865997   .3567258     1.08   0.314    -.4569229    1.230122
       _cons |  -.2498393   .2893364    -0.86   0.416    -.9340112    .4343326
------------------------------------------------------------------------------

. esttab, drop(2.dummy)

----------------------------
                      (1)   
                        y   
----------------------------
x                  0.0332   
                   (0.19)   

1.dummy                 0   
                      (.)   

_cons              -0.250   
                  (-0.86)   
----------------------------
N                      10   
----------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

. esttab, drop(1.dummy 2.dummy)

----------------------------
                      (1)   
                        y   
----------------------------
x                  0.0332   
                   (0.19)   

_cons              -0.250   
                  (-0.86)   
----------------------------
N                      10   
----------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

.

[Magnus Schauf]

You are right, thanks a lot!

[Joro Kolev]

And in related good news, it seems that Stata accepts wildcards, so you do not need to explicitly refer to the variables you want to drop. (Which will be a pain if you have many industries.)

. esttab, drop(*.dummy)

----------------------------
                      (1)   
                        y   
----------------------------
x                  0.0332   
                   (0.19)   

_cons              -0.250   
                  (-0.86)   
----------------------------
N                      10   
----------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

[Sonnen Blume]

And in related good news, it seems that Stata accepts wildcards, so you do not need to explicitly refer to the variables you want to drop. (Which will be a pain if you have many industries.)

. esttab, drop(*.dummy)

----------------------------
(1)
y
----------------------------
x 0.0332
(0.19)

_cons -0.250
(-0.86)
----------------------------
N 10
----------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

Thanks for the code!

[Nanna Lauritz]

I am running into a problem with this specific pain. How do I find out what stata called the variable?
I have a variable called "party_withmiss" and in the regression I use it as i.party_withmiss. I do not want this data in my estout regression output but I can't figure out how to drop the categorical variables. So how do I see what stata called the variables? I could do it with some of the labels by using shorthand of the label (like CDU*) - but that didn't work for all of the categories within the variable. Is there a rule of thumb?

[Andrew Musau]

Most estimation commands have an option -coeflegend- that allows you to see how coefficients are named, e.g.,

xtreg y x1 ... xn, fe robust coeflegend

Usually, factor variables have the format #.varname, where "#" is a numerical value representing the level. Therefore, for "i.party_withmiss", you'd need

drop(*.party_withmiss)