When calculating the risk ratio for factors set on a nominal scale in JMP, the results differ from those in SAS.
For nominal factors, the parameter estimation methods differ between JMP and SAS, so the hazard ratios calculated by exponentiating the parameter estimates differ.
To calculate the risk ratio of a target category when the risk ratio of a reference category is set to 1 in JMP, you must change the nominal scale to a continuous scale and perform the analysis.
More generally, we recommend conducting analyses using dummy variables for the variables of interest.
You can create a column consisting of 0/1 values by going to Columns > Utilities > Create Indicator Variable.
Also, while SAS allows you to specify any censoring value, JMP requires you to specify 1 (or another non-zero value) as the censoring value. Make sure the censoring values you specify in both analyses are consistent.
Confidence intervals for risk ratios (hazard ratios)
SAS uses Wald confidence intervals, while JMP uses likelihood-based confidence intervals (profile likelihood confidence intervals). (Note: In JMP 14.1 and later versions, Wald confidence intervals are used for confidence intervals of risk ratios (hazard ratios) for qualitative variables.) For this reason, the confidence intervals for risk ratios generally differ between JMP and SAS.
FAQ # 3004
