Prepared by Prajitha Nair
If frequency of the counts in any one of the cell is zero then the odds ratio cannot be computed?
Consider the example:
data gene;
input cat $7. genstatus $4. count;
cards;
case abn 119
control abn 0
case nor 317
control nor 319
;
run;
ods output RelativeRisks=relrisk(where=(studytype="Case-Control (Odds Ratio)"));
proc freq data=gene ;
tables cat*genstatus/relrisk ;
weight count;
run;
ods output close;
The output will look as follows:

This can be resolved as follows:
ods output CommonRelRisks=cmnrelrisk;
proc freq data=gene ;
tables cat*genstatus/relrisk cmh ;
weight count;
run;
ods output close;
The output will be produced as:

data cmnrelrisk_1(keep=Value LowerCL UpperCL);
set cmnrelrisk;
if studytype eq " (Odds Ratio)" then if method eq "Logit **" and value ne . then output;
else if studytype="Case-Control" then if method eq "Mantel-Haenszel" and value not in (.,0.0000) then output;
format LowerCL UpperCL ODDSR8.3 value ODDSR8.3;
run;
The cmh option will help to tackle the problem. The logit method will add 0.5 in all the cells of the 2 x 2 cross tabulation table where zero is in anyone of the cell and then odds ratio is evaluated.
Thus the odds ratio obtained is 240.506.
No comments:
Post a Comment