st: Bootstrapping with 'bca' option will not give BCa confidence interval

[Philip Jones <pjones8@uwo.ca>]

Hello,

I am trying to use bootstrapping to calculate the difference between two
medians, as in the following program. Unfortunately, even though the code
works correctly, I cannot seem to get BCa confidence intervals, although I can
get every other type of confidence interval.

Note that this problem seems to be the same as experienced in the following
Statalist discussion from 2005
(http://www.stata.com/statalist/archive/2005-12/msg00094.html)
but it doesn't seem to have been resolved as far as I can see.

My code is:

------------
capture program drop mydiff3
program mydiff3, rclass

version 12
gettoken number 0 : 0, parse(" ,")
gettoken grp 0 : 0, parse(" ,")

     
     quietly summarize `number' if `grp'==0, detail
     local median_group_hi = r(p50)
     quietly summarize `number' if `grp'==1, detail
     local median_group_lo = r(p50)
     return scalar difference = `median_group_lo'-`median_group_hi'

end


bootstrap r(difference), bca reps(1000) seed(12345) nodots saving(bs, replace) ///
nowarn: mydiff3 dose group
estat bootstrap, all
------------

This is used on a dataset with real numbers as the "number" variable, and an
indicator (0/1) variable ("grp").

I would greatly appreciate any help!

Thanks,

Phil
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/