How does one calculate Cohen's d and confidence intervals after logit in Stata? How does one calculate Cohen's d and confidence intervals after logit in Stata?
 A: Given the current issues with the user-written sizefx command for Stata that chl and I have uncovered, here's an alternative way of doing things in Stata using the user-written metan command. This is designed for meta-analysis of study-level summary data so you need to enter the means, sds and ns (which i've taken from chl's answer) in one row, then run a 'meta-analysis' on this single study (so the pooled SMD is the same as the SMD for the single study, and heterogeneity measures are undefined):. 

input mean0 sd0 n0 mean1 sd1 n1

         mean0        sd0         n0      mean1        sd1         n1
  1. 23.66154  5.584522  130  22.30508  4.511496        59
  2. end

. list

     +------------------------------------------------------+
     |    mean0        sd0    n0      mean1        sd1   n1 |
     |------------------------------------------------------|
  1. | 23.66154   5.584522   130   22.30508   4.511496   59 |
     +------------------------------------------------------+

. metan n1 mean1 sd1 n0 mean0 sd0, cohen

           Study     |     SMD   [95% Conf. Interval]     % Weight
---------------------+---------------------------------------------------
1                    | -0.257      -0.566     0.052        100.00
---------------------+---------------------------------------------------
I-V pooled SMD       | -0.257      -0.566     0.052        100.00
---------------------+---------------------------------------------------

  Heterogeneity chi-squared =   0.00 (d.f. = 0) p =    .
  I-squared (variation in SMD attributable to heterogeneity) =     .%

  Test of SMD=0 : z=   1.63 p = 0.103

The resulting point estimate and CI for Cohen's d agree with those that chl calculated by hand and using ci.smd() from R's MBESS library.
I'm less clear if this is entirely answering the original question, which asked to calculate Cohen's d after logit. But Cohen's d is one way of calculating the standardized mean difference, which I thought applied only to a continuous outcome measured in two groups. Perhaps the questioner could clarify.
