Summary Statistics

[Thomas Yes]

Hi,
I would like to do summary statistics (mean, st. dv., min, max, number of observations) for a single metric variable (income). I want to do it for the whole sample, and then for 2 binary dummy variables (full_time_employed and gender), so having 5 columns in total. Also, i need to export this to Word via estout. Really struggling with this at the moment. Anyone knows how to do this?

[Daniel Schaefer]

Hi Thomas, welcome to the forum. Suppose you have data that looks something like this:

clear
set obs 100
gen income = runiformint(5, 100) * 1000
gen full_time_employed = runiformint(0, 1)
gen gender = runiformint(0, 1)
label define empl 1 "employed" 0 "unemployed"
label values full_time_employed empl
label define genl 1 "women" 0 "men"
label values gender genl

Probably the most straightforward way to do this is with the -table- command.

table(full_time_employed gender), statistic(mean income) statistic(sd income) ///
    statistic(min income) statistic(max income) statistic(count income)

But that doesn't quite get you what you appear to be asking for because you don't want the multidimensional nesting structure.

-----------------------------------------------------------------------------------------------------------------
                   |      Mean   Standard deviation   Minimum value   Maximum value   Number of nonmissing values
-------------------+---------------------------------------------------------------------------------------------
full_time_employed |                                                                                             
  unemployed       |                                                                                             
    gender         |                                                                                             
      men          |  58538.46             31156.68            5000           99000                            26
      women        |  49692.31             26729.41            5000           90000                            26
      Total        |  54115.38             29086.52            5000           99000                            52
  employed         |                                                                                             
    gender         |                                                                                             
      men          |  51458.33             27741.01            5000           99000                            24
      women        |     51750             29266.76            9000           98000                            24
      Total        |  51604.17              28209.5            5000           99000                            48
  Total            |                                                                                             
    gender         |                                                                                             
      men          |     55140             29483.31            5000           99000                            50
      women        |     50680             27706.51            5000           98000                            50
      Total        |     52910             28551.95            5000           99000                           100
-----------------------------------------------------------------------------------------------------------------

I think this code gives you something more like what you are picturing:

gen employed = income if full_time_employed
gen unemployed = income if !full_time_employed
gen women = income if gender
gen men = income if !gender
table var result, statistic(mean employed unemployed women men income) ///
                  statistic(sd employed unemployed women men income) /// 
                  statistic(min employed unemployed women men income) ///
                  statistic(max employed unemployed women men income) ///
                  statistic(count employed unemployed women men income)

Which gives the following table of average incomes:

---------------------------------------------------------------------------------------------------------
           |      Mean   Standard deviation   Minimum value   Maximum value   Number of nonmissing values
-----------+---------------------------------------------------------------------------------------------
employed   |  51329.83             27748.62            5000          100000                           476
unemployed |  52940.84             27404.25            5000          100000                           524
women      |  51323.59             27465.68            5000          100000                           513
men        |  53069.82             27672.53            5000          100000                           487
income     |     52174             27566.64            5000          100000                         1,000
---------------------------------------------------------------------------------------------------------

That second solution might be improved by manually creating your own tags and dimensions using -collect- rather than creating new variables as I do, but the above seamed a bit simpler and easier to understand for someone new to tables. You may want to change the income variable name or the label in the table to "total". In any case, you should probably export the table to word with -putdocx- rather than -estout-.

putdocx begin
putdocx collect
putdocx save results.docx, replace

Finally, I would strongly recommend you read through the documentation for -table-, -collect-, and -putdox-. Automatically creating and exporting tables can get complicated. You probably want to familiarize yourself with the documentation.

[Jeff Pitblado (StataCorp)]

Daniel provides a solution facilitated by generating variables for each sample grouping.

Here is an alternative that builds a separate collection for each grouping, then combines them to produce a single table. Using the data from Daniel's example:

* summary statistics for the entire sample
table,  statistic(mean income)          ///
        statistic(sd income)            ///
        statistic(min income)           ///
        statistic(max income)           ///
        statistic(count income)         ///
        name(Total)
collect addtags Total[_hide]

* summary statistics within levels of full_time_employed
table (full_time_employed),             ///
        statistic(mean income)          ///
        statistic(sd income)            ///
        statistic(min income)           ///
        statistic(max income)           ///
        statistic(count income)         ///
        nototals                        ///
        name(full_time_employed)

* summary statistics within levels of gender
table (gender), ///
        statistic(mean income)          ///
        statistic(sd income)            ///
        statistic(min income)           ///
        statistic(max income)           ///
        statistic(count income)         ///
        nototals                        ///
        name(gender)

* combine these collections into a new collection named 'all'
collect combine all = full_time_employed gender Total

* add a title
collect title "Table: Summary of Income"

* shorten some result labels
collect label levels result     ///
        sd "Std. dev."          ///
        min "Minimum"           ///
        max "Maximum"           ///
        count "N", modify

* arrange the collected results into a table
collect layout (full_time_employed gender Total) (result)

Here is the resulting table.

Table: Summary of Income
--------------------------------------------------------------------
                   |      Mean   Std. dev.   Minimum   Maximum     N
-------------------+------------------------------------------------
full_time_employed |
  unemployed       |     56880     28722.2      8000    100000    50
  employed         |     54480     28382.4      7000    100000    50
gender             |
  men              |  58319.15    30107.85      8000    100000    47
  women            |  53339.62    26934.35      7000    100000    53
Total              |     55680    28433.83      7000    100000   100
--------------------------------------------------------------------

[Thomas Yes]

Thanks, that helped a lot!