Cox proportional test

[kabali bwogi]

my proportional-hazard assumption test based on Global test derived from Schoenfeld residuals was significant ----------------------------------------------------------------
| chi2 df Prob>chi2
------------+---------------------------------------------------
global test | 183.92 17 0.0000
----------------------------------------------------------------

what should i do????

[Maxence Morlet]

You know have to resort to parametric regression, see

help streg

Depending on your model, you may want to introduce a frailty parameter.

[Carlo Lazzaro]

Kabali:
you can stratify on the variable violating the proportional-hazard assumption and see what happens.
See, for more details, -stcox PHassumption tests-, Example 1, Stata .podf manual.

[Maxence Morlet]

Carlo is right, I entirely forgot about that. Please correct me if I'm wrong, but I believe you can also specify your predictor as a time-varing covariate.

[Carlo Lazzaro]

Maxence is right. This issue is covered at page 451-460 in the pivotal textbook http://gseacademic.harvard.edu/~alda/ .

[David Radwin]

Regarding these approaches, other citations that I found helpful are on pp. H-21-H-23 of 2008/18 Baccalaureate and Beyond Longitudinal Study (B&B:08/18) Data File Documentation.