Profile -likelihood CIs for risk ratios

[winfriedkoch0]

Proportional Hazards Fit provides profile likelihood based confidence intervals for the Parameter Estimates. Please provide this type of CIs also for the Risk Ratios; this is important because the risk ratios are the usual measure of effect size in survival analysis.

[SamGardner]

@winfriedkoch0 thank you for this new feature idea.  We are evaluating this idea as a possible future improvement to JMP.

[SamGardner]

@winfriedkoch0 we have investigated this request, and I wanted to let you know that we do calculate and report Wald confidence intervals for risk ratios (hazard ratios) in Fit Proportional Hazards.  These confidence intervals use normal approximations for the confidence limits.  Does this meet your needs?
( output from example data VA Lung Cancer)

Risk Ratios

Unit Risk Ratios

Per unit change in regressor

Term	Risk Ratio	Lower 95%	Upper 95%	Reciprocal
Age	0.991487	0.9739	1.010136	1.0085861
Diag Time	0.999908	0.979971	1.016244	1.000092
KPS	0.967905	0.957488	0.978403	1.0331596

Range Risk Ratios

Per change in regressor over entire range

Term	Risk Ratio	Lower 95%	Upper 95%	Reciprocal
Age	0.669099	0.288517	1.606371	1.4945466
Diag Time	0.992119	0.175518	3.998028	1.0079435
KPS	0.05484	0.020935	0.143244	18.23482

Risk Ratios for Cell Type

Level1	/Level2	Risk Ratio	Prob>Chisq	Lower 95%	Upper 95%
Large	Adeno	0.4544481	0.0092*	0.2511038	0.8224612
Small	Adeno	0.7176217	0.2286	0.4181316	1.2316242
Squamous	Adeno	0.3047391	<.0001*	0.1690124	0.5494621
Small	Large	1.579106	0.0862	0.9370433	2.6611104
Squamous	Large	0.6705696	0.1574	0.3853373	1.166935
Adeno	Large	2.2004712	0.0092*	1.2158629	3.9824176
Squamous	Small	0.4246514	0.0019*	0.2476225	0.7282408
Adeno	Small	1.3934918	0.2286	0.811936	2.3915917
Large	Small	0.6332697	0.0862	0.375783	1.0671865
Adeno	Squamous	3.2814957	<.0001*	1.8199618	5.916725
Large	Squamous	1.4912695	0.1574	0.8569458	2.5951289
Small	Squamous	2.3548727	0.0019*	1.3731721	4.0384051

Risk Ratios for Treatment

Level1	/Level2	Risk Ratio	Prob>Chisq	Lower 95%	Upper 95%
Test	Standard	1.3363418	0.1617	0.8903074	2.0058347
Standard	Test	0.7483115	0.1617	0.4985456	1.1232076

Risk Ratios for Prior

Level1	/Level2	Risk Ratio	Prob>Chisq	Lower 95%	Upper 95%
Yes	No	1.0750063	0.7554	0.6820551	1.6943478
No	Yes	0.9302271	0.7554	0.5901976	1.4661572

[SamGardner]

Based on additional communication, the Wald confidence intervals are sufficient to meet this need.  Thank you for your suggestion and continued engagement with JMP!