Does the variable of interest (i.e. the novel metric) change over time for each specific individual?

We only calculated the novel metric once, although the phenomenon the metric identifies can change over time.

Dear Dr. Harini Subramanian , Thank you for your reply. I understand that the value of the metric you calculated can change over time for an individual, just as the body mass index changes.

Try using a flexible parametric model with stpm2 in Stata, or simply use an AFT. Start with stpm2 if the model converges well you’re good. If not, AFT requires a smaller sample size.

You could increase the sample size, which will increase power and nullify the dependence on the PH assumption :)

hi Harini Subramanian For a small sample (n < 100) where the proportional hazards assumption is violated, consider Accelerated Failure Time (AFT), it works well with small samples.