Date:         Thu, 4 Mar 2010 08:26:46 -0800
Reply-To:     "J.k" <cakheil@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "J.k" <cakheil@GMAIL.COM>
Organization: http://groups.google.com
Subject:      PHREG w/ strata question
Comments: To: sas-l@uga.edu
Content-Type: text/plain; charset=ISO-8859-1

I've run into a question regarding how PHREG functions when using the
strata statement to generate risk sets for a conditional logistic
regression modeling match case control data (1:N). Each strata group
originally had at least one case and one control. However after
applying an exclusion criteria some of the risk sets lost their only
case or only control. These groups with no event or censored values
seem to be still included by PHREG. Should PHREG have removed them
automatically?  Do I need to remove them prior to calling the
procedure or should they be left in the model?  It seems to me that
all 35 of the groups should be excluded but I'm more of a programmer
than a statistician. Any comments would be greatly appreciated.
Thanks!

Model Information

Data Set                 WORK.X_MASTERFILE
Dependent Variable       status
Censoring Variable       CC_STATUS             Case/control status
Censoring Value(s)       0
Ties Handling            BRESLOW


Number of Observations Read        7510
Number of Observations Used        7510

        Summary of the Number of Event and Censored Values


Percent
Pattern       Total       Event    Censored    Censored      Freq

      1           1           0           1
100.00          4
      2           1           1           0
0.00          1
      3           2           0           2
100.00        30
      4           2           1           1
50.00      520
      5           3           1           2                   66.67
2135