Proportional Hazards Fit, Confidence Intervals for Risk Ratios

[#1]

The Proportional Hazards Fit Procedure in JMP provides Risk Ratios together with Wald Test Type Confidence intervals. Is there a possibility to get profile likelihood based CIs as these are usually smaller compared to Wald based CIs?

[#2]

Hi,

I've just searched the help documentation for Fit Proportional Hazards and, yes, it appears that you can get profile likelihood confidence intervals:

"Likelihood Confidence Intervals

Specifies the type of confidence intervals shown in the Parameter Estimates table for each parameter. When this option is selected, a profile likelihood confidence interval appears. Otherwise, a Wald interval is shown. In the report, the interval type is noted below the Parameter Estimates table. This option is on by default when the computational time for the profile likelihood confidence intervals is not large."

You can find the option in the red triangle menu for Proportional Hazards Fit:

Phil_Kay_0-1646387935499.png

I hope this helps.

Phil

[#3]

Dear Phil,

Thanks for your fast feedback,

From my understanding your recommendation is true and applicable for the section "Parameter Estimates" in the output but not for the section "Risk Ratios".

When I apply your recommended settings, no change is visible for the confidence intervals of the Risk Ratios.

The note under Risk ratios is always: " Normal approximations used for ratio CIs"

Is it possible to calculate the risk ratios from the Parameter Estimates?

[#4]

Okay. Yes, sorry I missed that.

It does not look like that is possible.

I suggest that you add a request in the JMP Wish List.

Regards,

Phil