variable means and p-values for treatment and control group

[Pablo Juri]

Hi everyone,

for my assignment I need to test my regression discontinuity sample for balance. In order to do so I want to compare various means of baseline characteristics for treatment and control group.
I managed to output the means and compare them in a table via outreg2 but now I struggle to get the p-value to check for significant differences.

My code so far was:

global variables_baseline "my variables"
estpost sum $variables_baseline
outreg2 using baseline if survey==2016 & in_bw==1 & treatment==0, sum keep($variables_baseline) stats(mean) ct(Control Group) title (Characteristics at baseline) label dec(3) replace
outreg2 using baseline if survey==2016 & in_bw==1 & treatment==1, sum keep($variables_baseline) stats(mean) ct(Treatment Group) label dec(3) append word

Is it possible to get the needed p-values for significant differences in the third column?

Thanks a lot in advance!

[George Ford]

Imbens and Wooldridge (among others) do not recommend t-test since it depends on sample size (so trivially small differences may be significant in large samples). Standardized Differences is often recommended. It is akin to a t-test but excludes sample size (large is >0.25, though some say 0.10 in epidemiology). The command -covbal- does the trick. I think you can "ssc install covbal".

If you insist, then use ttest and scroll through your X.

foreach var in $variables_baseline {
    qui ttest `var' if survey==2016 & in_bw==1, by(treatment)
    di "`var'" _col(20) "t = " %7.3f r(t) _col(40) "pr = " %5.3f r(p)
}

[Pablo Juri]

Worked!! Thank you very much!