Formatting the by() variable in tabstat

[Rafael Copat]

Hello,

I am working with a database structured at the firm-year-month level. I am using --tabstat-- to create some summary statistics by the time variable. If I use the option "columns(statistics)", --tabstat-- retains the formatting of the time variable (%tm). However, if I use "columns(variables)", it does not retain the formatting. Is this behavior expected or a bug? Is there a way to format the time variable as %tm in the last table (please, see formating of the "year_month" variable in the first three tables versus the last one in the code below)? If not, can I achieve what I want using --table--? I tried --table--, but was facing the same issue.

Any thoughts are appreciated.

version 17
cls
clear all

set seed 928502853
set obs 24
gen year_month = tm(2010,Jan) + _n - 1
format %tm year_month

expand 30
bysort year_month: gen firm_id = _n

foreach num of numlist 1(1)3 {
    gen var_`num' = rnormal()
}

tabstat var_1, by(year_month) statistics(mean sd min p25 p50 p75 max count) columns(statistics) format(%9.2fc)
tabstat var_2, by(year_month) statistics(mean sd min p25 p50 p75 max count) columns(statistics) format(%9.2fc)
tabstat var_3, by(year_month) statistics(mean sd min p25 p50 p75 max count) columns(statistics) format(%9.2fc)

tabstat var_1 var_2 var_3, by(year_month) statistics(mean) columns(variables) format(%9.2fc)

[Justin Niakamal]

I've found labmask (from SSC written by Nick Cox) is useful for instances such as these. You can also specify your desired format in your string() call.

cls
clear all

set seed 928502853
set obs 24
gen year_month = tm(2010,Jan) + _n - 1
format %tm year_month

expand 30
bysort year_month: gen firm_id = _n

foreach num of numlist 1(1)3 {
    gen var_`num' = rnormal()
}


gen month = string(year_month, "%tm")
labmask year_month, val(month)
drop month

tabstat var_1, by(year_month) statistics(mean sd min p25 p50 p75 max count) columns(statistics) format(%9.2fc)
tabstat var_2, by(year_month) statistics(mean sd min p25 p50 p75 max count) columns(statistics) format(%9.2fc)
tabstat var_3, by(year_month) statistics(mean sd min p25 p50 p75 max count) columns(statistics) format(%9.2fc)

tabstat var_1 var_2 var_3, by(year_month) statistics(mean) columns(variables) format(%9.2fc)

year_mo~h |      mean        sd       min       p25       p50       p75       max         N
----------+--------------------------------------------------------------------------------
   2010m1 |      0.02      1.05     -2.50     -0.91      0.26      0.93      1.46     30.00
   2010m2 |     -0.06      1.13     -3.33     -0.82      0.02      0.68      1.94     30.00
   2010m3 |      0.04      1.02     -2.10     -0.52     -0.08      0.46      3.13     30.00
   2010m4 |      0.11      0.92     -1.80     -0.55      0.22      0.66      1.83     30.00
   2010m5 |      0.58      0.85     -1.25     -0.07      0.43      1.20      2.20     30.00
   2010m6 |     -0.11      0.88     -2.04     -0.66     -0.13      0.39      1.49     30.00
   2010m7 |      0.21      0.95     -2.21     -0.57      0.27      0.72      2.55     30.00
   2010m8 |     -0.13      1.15     -2.65     -0.94     -0.02      0.54      2.01     30.00
   2010m9 |      0.16      0.80     -1.65     -0.34      0.10      0.69      1.90     30.00
  2010m10 |      0.14      0.88     -1.63     -0.57      0.19      0.85      1.57     30.00
  2010m11 |     -0.42      1.27     -2.16     -1.36     -0.72      0.52      2.57     30.00
  2010m12 |      0.29      1.01     -1.54     -0.60      0.39      1.03      2.51     30.00
   2011m1 |     -0.00      0.97     -2.67     -0.48     -0.03      0.57      2.18     30.00
   2011m2 |      0.26      1.17     -2.05     -0.52      0.19      0.98      2.94     30.00
   2011m3 |     -0.19      1.41     -2.87     -1.25     -0.33      0.92      3.39     30.00
   2011m4 |     -0.13      0.67     -1.37     -0.45     -0.18      0.31      1.17     30.00
   2011m5 |      0.11      0.88     -1.52     -0.45      0.11      0.48      1.95     30.00
   2011m6 |      0.08      1.20     -2.41     -0.36      0.25      0.86      2.28     30.00
   2011m7 |     -0.12      1.28     -2.72     -1.17      0.07      0.76      2.59     30.00
   2011m8 |     -0.31      0.91     -2.22     -0.97     -0.47      0.15      1.92     30.00
   2011m9 |      0.08      1.10     -1.88     -0.66     -0.03      0.82      2.20     30.00
  2011m10 |      0.08      1.09     -2.02     -0.72      0.28      0.83      2.15     30.00
  2011m11 |      0.12      0.98     -1.46     -0.51     -0.10      0.72      2.24     30.00
  2011m12 |      0.01      0.80     -1.47     -0.57     -0.03      0.57      1.62     30.00
----------+--------------------------------------------------------------------------------
    Total |      0.04      1.03     -3.33     -0.62      0.05      0.72      3.39    720.00
-------------------------------------------------------------------------------------------

[Rafael Copat]

Hi Justin,

Thank you very much for your response. I had tried multiple workarounds without success. It seems the way to go is to encode the time variable, and --labmask-- makes things much easier! Thank you, and thanks to Nick Cox for --labmask--!

For those that are having the same issue, --labmask-- is not on SSC, but on the Stata Journal's webpage. It can be installed with the following command:

net install "http://www.stata-journal.com/software/sj8-2/gr0034"

Finally, just to be clear, the output that Justin reports above can be obtained without the use of --encode-- or --labmask--. It is the last --tabstat-- command that requires one of the two commands.

tabstat var_1 var_2 var_3, by(year_month) statistics(mean) columns(variables) format(%9.2fc)

Justin Niakamal, thank you one more time. Regards,

Rafael

[Nick Cox]

Glad labmask is useful. On SSC it is part of the labutil package, but it’s better cited in terms of a Stata Journal link, given the associated 2008 paper.