Relative Risk for Small Sample

[taishayla]

Any suggestions for which procedure to use for relative risk in which the sample size is small and two-by-two table has 0 values? My table, top to bottom/left to right: 10/0/1/3. I've used the relrisk option with proc freq. Wasn't sure if it may be best to use the logistic model and if the firth option would best handle this. Suggestions are welcomed!

[Reeza]

PROC FREQ with the EXACT options specified?

[taishayla]

Yes

[StatDave]

The results from EXACT RELRISK; in PROC FREQ are what you want. That provides exact results. The Firth method is not an exact method - it just uses a penalized likelihood.