Chelsea,
I haven't used regular expressions in a bit, so someone should correct me if I'm wrong, but I think the problem you mention would be solved by replacing Tom's first line of code with:
gen group = (regexm(HIV, "N[\.I]*P"))
The expression in the brackets tells Stata to ignore "."s and "I"s when it looks for an N followed by a P. For more info on using regular expressions, check here: http://www.stata.com/support/faqs/data/regex.html. -help regexm- will also be useful.
Also, I don't know if you have any cases where someone is indeterminate or missing in every period. If you have any such cases, I think Tom's code will code those as group 3, which does not seem appropriate. You may want to add a line of code as follows:
replace group = 4 if !regexm(HIV, "N") & !regexm(HIV, "P")
You will also want value labels for your variable. The following code should -label- your group variable.
lab define grouplab 1 "incident seroconverter" 2 "prevalent positive" ///
3 "consistently seronegative" 4 "missing/indeterminate"
lab val group grouplab
Lastly, I'd like to warn you that I've had some trouble with the way that Stata's -regexm- function deals with missing values. If you have any truly missing values (i.e. ""), I would carefully check to make sure that the -regexm- function is dealing with them in the right way. See this thread for the problem I had:
http://www.stata.com/statalist/archive/2008-10/msg00935.html
Hope this helps
Howie
Howie Lempel
Research Assistant
The Brookings Institution | Economic Studies
1775 Massachusetts Ave NW | Washington DC 20036
[email protected] | p: (202) 238-3576
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Polis, Chelsea B.
Sent: Monday, November 24, 2008 8:49 PM
To: [email protected]
Subject: RE: st: Categorizing HIV status using a series of string variables
Many thanks, Tom and Howie!
Tom: Your solution worked beautifully, except for one tiny thing. I got a few people who weren't assigned to one of the three groups, and their codes all had one thing in common...an "indeterminate" test between their negative and positive tests:
hiv
......NI..PP
.....NIP....
......NI...P
I can very easily just recode these people to be incident seroconverters, but I wonder if there is an easy fix for the code that would do this automatically?
Howie: Many thanks for the explanation...that clears up a lot of my confusion!
BTW: my apologies if this post ends up in the wrong spot - I'm still trying to understand how to reply to individual postings when I receive Statalist in digest form...I'm hoping that slapping a "RE:" in front of the subject line I wish to respond to will allow me to do that, but I couldn't find information in the FAQ on specifically how to do this.
Cheers,
Chelsea
*
* 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/
*
* 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/
