st: Treatment of odds ratios for tables with a 0 cell frequency

[Mike Lacy <Michael.Lacy@colostate.edu>]

I just had occasion to notice that commands that produce confidence 
limits for odds ratios in Stata (e.g., cc, mhodds) decline to 
calculate exact confidence limits when there is a zero cell in the 
table.  For example,
 -cci 10 20 30 0- yields "Note: exact confidence levels not possible 
with zero count cells."  To my knowledge, this is not the practice in 
other programs that calculate exact CIs for odds ratios, such as 
StatXact: they will set the lower limit to 0 and calculate an exact 
upper limit, or set the upper limit as undefined and calculate an 
exact lower limit.  Since there is no particular algorithmic 
difficulty in doing this calculation (in contrast to the error 
message), does anyone have an idea of why the choice not to calculate 
these limits has been implemented in Stata?  I personally think it is 
useful to know, for example, what the lower confidence limit is on 
the OR even when the observed value is infinite due to a zero cell.

=-=-=-=-=-=-=-=-=-=-=-=-=
Mike Lacy
Fort Collins CO USA
(970) 491-6721 office


*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/