Export T-test to Latex

[Chibo Dike]

Please, I would like export test results from stata to latex. Here's what I want to do.

sysuse auto

test of the difference between the variables 'price' and 'turn'

ttest price=turn

which yields

Paired t test
------------------------------------------------------------------------------
Variable |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
---------+--------------------------------------------------------------------
   price |      74    6165.257    342.8719    2949.496    5481.914      6848.6
    turn |      74    39.64865    .5114145    4.399354     38.6294     40.6679
---------+--------------------------------------------------------------------
    diff |      74    6125.608    342.7139    2948.137     5442.58    6808.636
------------------------------------------------------------------------------
     mean(diff) = mean(price - turn)                              t =  17.8738
 Ho: mean(diff) = 0                              degrees of freedom =       73

 Ha: mean(diff) < 0           Ha: mean(diff) != 0           Ha: mean(diff) > 0
 Pr(T < t) = 1.0000         Pr(|T| > |t|) = 0.0000          Pr(T > t) = 0.0000

Please, how can I export the last row ('diff') to latex? I am aware of estpost ttest, but it requires specifying a group variable which does not apply to my case. I am also aware of asdoc, but it's for exporting to word. I have resorted to manually coping and pasting results from stata into a table in latex but that's tedious especially when many variables are involved and one needs to run some regression several times and results change.

[Andrew Musau]

estout is from the Stata Journal. Consider this:

sysuse auto
ttest price=turn
preserve
expand 2, g(new)
gen diff= cond(new, turn, price)
estpost ttest diff, by(new)
esttab
restore

Res.:

. ttest price=turn

Paired t test
------------------------------------------------------------------------------
Variable |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
---------+--------------------------------------------------------------------
   price |      74    6165.257    342.8719    2949.496    5481.914      6848.6
    turn |      74    39.64865    .5114145    4.399354     38.6294     40.6679
---------+--------------------------------------------------------------------
    diff |      74    6125.608    342.7139    2948.137     5442.58    6808.636
------------------------------------------------------------------------------
     mean(diff) = mean(price - turn)                              t =  17.8738
 Ho: mean(diff) = 0                              degrees of freedom =       73

 Ha: mean(diff) < 0           Ha: mean(diff) != 0           Ha: mean(diff) > 0
 Pr(T < t) = 1.0000         Pr(|T| > |t|) = 0.0000          Pr(T > t) = 0.0000


. ttest diff, by(new)

Two-sample t test with equal variances
------------------------------------------------------------------------------
   Group |     Obs        Mean    Std. Err.   Std. Dev.   [95% Conf. Interval]
---------+--------------------------------------------------------------------
       0 |      74    6165.257    342.8719    2949.496    5481.914      6848.6
       1 |      74    39.64865    .5114145    4.399354     38.6294     40.6679
---------+--------------------------------------------------------------------
combined |     148    3102.453    304.9675     3710.09    2499.766     3705.14
---------+--------------------------------------------------------------------
    diff |            6125.608    342.8723                5447.974    6803.242
------------------------------------------------------------------------------
    diff = mean(0) - mean(1)                                      t =  17.8656
Ho: diff = 0                                     degrees of freedom =      146

    Ha: diff < 0                 Ha: diff != 0                 Ha: diff > 0
 Pr(T < t) = 1.0000         Pr(|T| > |t|) = 0.0000          Pr(T > t) = 0.0000


. esttab

----------------------------
                      (1)  
                            
----------------------------
diff               6125.6***
                  (17.87)  
----------------------------
N                     148  
----------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001

[Chibo Dike]

Thanks Andrew. One more thing, please. Let's say groups 0 and 1 are actual and predicted values of some variable. Is it possible to reshape the table to the form below?:

                                 (1)                     (2)                       (3)
                        actual                    predicted          ATE=actual-predicted
        mean           6165.275           39.64865                         6125.6 (17.87)

Sorry the table alignment is poor. But the point is that under column (1) you have mean for actual, (2) for predicted, and (3) shows (1) - (2).

[Andrew Musau]

sysuse auto
ttest price=turn
preserve
expand 2, g(new)
gen diff= cond(new, turn, price)
estpost ttest diff, by(new)
esttab, cells(" b(fmt(3) pvalue(p) star) mu_1(fmt(3)) mu_2(fmt(3))" ///
t(par fmt(2))) collab("ATE" "actual" "predicted") nonumbers coeflab(diff "mean")
restore

Res.:

. esttab, cells(" b(fmt(3) pvalue(p) star) mu_1(fmt(3)) mu_2(fmt(3))" ///
> t(par fmt(2))) collab("ATE" "actual" "predicted") nonumbers coeflab(diff "mean")

------------------------------------------------------
                                                      
                      ATE          actual    predicted
------------------------------------------------------
mean             6125.608***     6165.257       39.649
                  (17.87)                             
------------------------------------------------------
N                     148                             
------------------------------------------------------

[Chibo Dike]

Thanks a lot Andrew. Please, how can I include the standard errors in brackets (as it usually appears in regression tables) under ATE, actual and predicted. I simply took out "t(par fmt(2)))" and put "se(par))" in the code you gave as follows:

esttab, cells(" b(fmt(3) pvalue(p) star) mu_1(fmt(3)) mu_2(fmt(3))" /// 
se(par)) collab("ATE" "actual" "predicted") nonumbers coeflab(diff "mean")

which gives

------------------------------------------------------
                                                      
                      ATE          actual    predicted
------------------------------------------------------
mean             6125.608***     6165.257       39.649
                (342.872)                             
------------------------------------------------------
N                     148                             
------------------------------------------------------

Please, could you show how to include the standard errors for actual and predicted?

[Andrew Musau]

sysuse auto
ttest price=turn
preserve
expand 2, g(new)
gen diff= cond(new, turn, price)
eststo m1: mean diff if !new
eststo m2: mean diff if new
eststo m3: estpost ttest diff, by(new)
esttab m1 m2 m3, se mlab(actual predicted ATE) coeflab(diff "mean")
restore

Res.:

 
. esttab m1 m2 m3, se mlab(actual predicted ATE) coeflab(diff "mean")

------------------------------------------------------------
                      (1)             (2)             (3)   
                   actual       predicted             ATE   
------------------------------------------------------------
mean               6165.3***        39.65***       6125.6***
                  (342.9)         (0.511)         (342.9)   
------------------------------------------------------------
N                      74              74             148   
------------------------------------------------------------
Standard errors in parentheses
* p<0.05, ** p<0.01, *** p<0.001

[Chibo Dike]

Thanks a lot. Sorry to drag this on. I want the stars to appear only on ATE. From the esttab documentation I see it's the stardrop or starkeep option. I've tried stardrop(actual predicted), starkeep(diff) among others but all result in errors.

[Andrew Musau]

You can use the option -nostar- to suppress all the starts. If you ask me, it does not make sense to show standard errors for all and then only show stars for some. Either just show the means for statistics without stars or be consistent if displaying standard errors.

[Chibo Dike]

Ok. The stars are a very minor issue. Thanks a lot for your help.

[Andrew Musau]

If you must, use the -substitute- option, but also change the note at the foot of the table so that the reader is not misled.

sysuse auto, clear
ttest price=turn
preserve
expand 2, g(new)
gen diff= cond(new, turn, price)
eststo m1: mean diff if !new
eststo m2: mean diff if new
eststo m3: estpost ttest diff, by(new)
esttab m1 m2 m3, se mlab(actual predicted ATE) coeflab(diff "mean") ///
substitute("39.65***" "39.65   "  "6165.3***"   "6165.3   ") ///
addnote("Significance stars only shown for ATE estimates")

Res.:

------------------------------------------------------------
                      (1)             (2)             (3)  
                   actual       predicted             ATE  
------------------------------------------------------------
mean               6165.3           39.65          6125.6***
                  (342.9)         (0.511)         (342.9)  
------------------------------------------------------------
N                      74              74             148  
------------------------------------------------------------
Standard errors in parentheses
Significance stars only shown for ATE estimates
* p<0.05, ** p<0.01, *** p<0.001

[Chibo Dike]

Perfect. Thanks a lot.

[Tiago Munhoz]

Thank you all for helpful tips. I have been looking for code to export summary tables (mean, SD) for categorical variables.
Any simple way to export “oneway price rep78, tab” results to Latex, excel or similar?
PS: As example here am assuming rep78 as a categorial variable.

[Jeff Pitblado (StataCorp)]

With the new table building features in Stata 17 you can design custom styles and labels for your table, save them for use with the table command, and then export your table to LaTeX.

Here is some Stata 17 code to almost reproduce Tiago's example above.

sysuse auto

* create custom labels and styles
* custom labels
collect label dim result "Summary of Price"
collect label levels result sd "Std. dev." frequency "Freq."
* start with with factory default styles
collect style clear
* usual borders between the table headers and table data
collect style cell border_block[corner], border(right bottom)
collect style cell border_block[row-header], border(right)
collect style cell border_block[column-header], border(bottom)
* border for "Total"
collect style cell cell[row-header]#rep78[.m], border(right)
collect style cell rep78[.m], border(top)
* numeric format
collect style cell result[mean sd], nformat(%9.2fc)
* header styles
collect style row stack, noindent length(11)
collect style column, dups(center)
* save label and styles to a single file, easy to reference
collect save mysumtab, replace

* compute summary statistics and specify our styles and labels
table rep78, ///
    stat(mean price) ///
    stat(sd price) ///
    stat(freq) ///
    style(mysumtab) ///
    label(mysumtab)

* export to LaTeX
collect export example.tex, replace
type example.tex

Here is a log of the above.

. sysuse auto
(1978 automobile data)

. 
. * create custom labels and styles
. * custom labels
. collect label dim result "Summary of Price"

. collect label levels result sd "Std. dev." frequency "Freq."

. * start with with factory default styles
. collect style clear

. * usual borders between the table headers and table data
. collect style cell border_block[corner], border(right bottom)
(dimension border_block not found)

. collect style cell border_block[row-header], border(right)
(level row-header of dimension border_block not found)

. collect style cell border_block[column-header], border(bottom)
(level column-header of dimension border_block not found)

. * border for "Total"
. collect style cell cell[row-header]#rep78[.m], border(right)
(dimension cell not found)
(dimension rep78 not found)

. collect style cell rep78[.m], border(top)

. * numeric format
. collect style cell result[mean sd], nformat(%9.2fc)
(dimension result not found)

. * header styles
. collect style row stack, noindent length(11)

. collect style column, dups(center)

. * save label and styles to a single file, easy to reference
. collect save mysumtab, replace
(collection default saved to file mysumtab.stjson)

. 
. * compute summary statistics and specify our styles and labels
. table rep78, ///
>         stat(mean price) ///
>         stat(sd price) ///
>         stat(freq) ///
>         style(mysumtab) ///
>         label(mysumtab)

            |     Summary of Price    
            |     Mean Std. dev. Freq.
------------+-------------------------
     Repair |                         
record 1978 |                         
          1 | 4,564.50    522.55     2
          2 | 5,967.62  3,579.36     8
          3 | 6,429.23  3,525.14    30
          4 | 6,071.50  1,709.61    18
          5 | 5,913.00  2,615.76    11
------------+-------------------------
      Total | 6,146.04  2,912.44    69

. 
. * export to LaTeX
. collect export example.tex, replace
(collection Table exported to file example.tex)

. type example.tex
\documentclass{article}
\usepackage{multirow}
\usepackage{amsmath}
\usepackage{ulem}
\usepackage[table]{xcolor}
\begin{document}
\begin{tabular}{llll}
\multicolumn{1}{r}{} &
  \multicolumn{3}{|c}{Summary of Price} \\
\multicolumn{1}{r}{} &
  \multicolumn{1}{|r}{Mean} &
  \multicolumn{1}{r}{Std. dev.} &
  \multicolumn{1}{r}{Freq.} \\
\cline{1-4}
\multicolumn{1}{r}{Repair} &
  \multicolumn{1}{|r}{} &
  \multicolumn{1}{r}{} &
  \multicolumn{1}{r}{} \\
\multicolumn{1}{r}{record 1978} &
  \multicolumn{1}{|r}{} &
  \multicolumn{1}{r}{} &
  \multicolumn{1}{r}{} \\
\multicolumn{1}{r}{1} &
  \multicolumn{1}{|r}{4,564.50} &
  \multicolumn{1}{r}{522.55} &
  \multicolumn{1}{r}{2} \\
\multicolumn{1}{r}{2} &
  \multicolumn{1}{|r}{5,967.62} &
  \multicolumn{1}{r}{3,579.36} &
  \multicolumn{1}{r}{8} \\
\multicolumn{1}{r}{3} &
  \multicolumn{1}{|r}{6,429.23} &
  \multicolumn{1}{r}{3,525.14} &
  \multicolumn{1}{r}{30} \\
\multicolumn{1}{r}{4} &
  \multicolumn{1}{|r}{6,071.50} &
  \multicolumn{1}{r}{1,709.61} &
  \multicolumn{1}{r}{18} \\
\multicolumn{1}{r}{5} &
  \multicolumn{1}{|r}{5,913.00} &
  \multicolumn{1}{r}{2,615.76} &
  \multicolumn{1}{r}{11} \\
\cline{1-4}
\multicolumn{1}{r}{Total} &
  \multicolumn{1}{|r}{6,146.04} &
  \multicolumn{1}{r}{2,912.44} &
  \multicolumn{1}{r}{69} \\
\end{tabular}
\end{document}

And here is the PDF result after running pdflatex example.

[Tiago Munhoz]

Thanks a lot, Jeff. Fantastic solution/example.
Unfortunately, I am using Stata 16.1 and cannot run it. Anyway, thanks for helping me.

[Andrew Musau]

Thanks Jeff Pitblado (StataCorp) for the nice illustration. I suspect that these external exporting routines will fall into disuse over time. The statistics in the requested table can be generated using standard egen functions. The following uses estout from SSC to export these. All you need to be changing in the code are the names of the variables (highlighted in red).

sysuse auto, clear
local byvar rep78
gen Mean= price
drop if missing(`byvar')
expand 2, g(new)
qui sum `byvar'
assert `byvar'!=99999
replace `byvar'= 99999 if new
bys `byvar': egen Std_Err=sd(Mean)
bys `byvar': egen Freq=count(Mean)
collapse Mean Std_Err Freq, by("`byvar'")
estpost tabstat  Mean Std_Err Freq, by("`byvar'")
esttab, cells("Mean Std_Err Freq") noobs nomtitle nonumber ///
collab(,lhs("`:var lab `byvar''")) drop(Total) ///
substitute("99999" "Total" "Std_Err" "Std. Err") varwidth(20)
*LaTeX Table
esttab, cells("Mean Std_Err Freq") noobs nomtitle nonumber ///
collab(,lhs("`:var lab `byvar''")) drop(Total) ///
substitute("99999" "Total" "Std_Err" "Std. Err") varwidth(20) tex

Res.:

.
. esttab, cells("Mean Std_Err Freq") noobs nomtitle nonumber ///
> collab(,lhs("`:var lab `byvar''")) drop(Total) ///
> substitute("99999" "Total" "Std_Err" "Std. Err") varwidth(20)

-----------------------------------------------------------
Repair Record 1978           Mean      Std. Err         Freq
-----------------------------------------------------------
1                          4564.5     522.5519            2
2                        5967.625     3579.357            8
3                        6429.233      3525.14           30
4                          6071.5     1709.608           18
5                            5913     2615.763           11
Total                    6146.043      2912.44           69
-----------------------------------------------------------

. *LaTeX Table
. esttab, cells("Mean Std_Err Freq") noobs nomtitle nonumber ///
> collab(,lhs("`:var lab `byvar''")) drop(Total) ///
> substitute("99999" "Total" "Std_Err" "Std. Err") varwidth(20) tex

{
\def\sym#1{\ifmmode^{#1}\else\(^{#1}\)\fi}
\begin{tabular}{l*{1}{ccc}}
\hline\hline
Repair Record 1978  &        Mean&     Std. Err&        Freq\\
\hline
1                   &      4564.5&    522.5519&           2\\
2                   &    5967.625&    3579.357&           8\\
3                   &    6429.233&     3525.14&          30\\
4                   &      6071.5&    1709.608&          18\\
5                   &        5913&    2615.763&          11\\
Total               &    6146.043&     2912.44&          69\\
\hline\hline
\end{tabular}
}