topic Re: How to include subjects when calculating commonrisdiff when subject present only in one treatgro in Statistical Procedures

How to include subjects when calculating commonrisdiff when subject present only in one treatgroup

knveraraju91 — Tue, 30 May 2023 15:46:26 GMT

I am calculating risk diff between treatment groups 1 and 2 by below pgm. The first record 'variable SS =1 'present only for trt=1. I think this record is omitted by freq pgm because i am getting same values( value,stderr,lowerci,and upperci) when i ran pgm after excluding first record. I need to consider the first record also in calculating riskdiff between trteatments. Please suggest me. Thank you

data one;
input trt res$ age ss ph $;
datalines;
1 cr 55 1 Y
1 cr 50 2 N
1 cri 55 2 Y
1 cri 50 2 N
1 cri 55 2 N
1 cr 50 2 N
1 cri 55 2 Y
1 cri 50 2 N
2 cri 55 2 Y
2 cri 50 2 N
2 cr 55 2 Y
2 cr 50 2 N
2 cri 55 2 Y
2 cr 50 2 Y
2 cri 55 2 N
2 cri 50 2 N
;

ods select none;
ods output CommonPdiff=rddds RiskDiffPlot=RiskDiffPlot;
 proc freq data = one;
                tables age*ss*ph* trt * res / riskdiff(common)  alpha = 0.05 ;
        run ;
ods select all;

Re: How to include subjects when calculating commonrisdiff when subject present only in one treatgro

SteveDenham — Tue, 30 May 2023 17:51:22 GMT

Since all records for trt 2 have ss=2, and all but one for trt 1 have ss=2, I would suggest rearranging the TABLES statement. If you truly need to look at the effect of ss, then put it next to last as in:

 tables age*ph* trt *ss* res / riskdiff(common)  alpha = 0.05 ;

Pardon the lined out part. It won't help to do anything like what I suggested there.

Because there are no records with ss=1 for any other cells in the crosstabulation, it always ends in a column associated with all zeroes for trt and res, and is excluded from riskdiff calculations. You may be able to use an EXACT statement in PROC GENMOD to include this record, but without doing a lot of digging, I don't know if that will give you what you need to calculate a riskdiff.

SteveDenham