topic Re: proc PHREG takes too long to assess proportional hazards assumption in Statistical Procedures

proc PHREG takes too long to assess proportional hazards assumption

akhmerova — Thu, 04 Mar 2021 19:54:37 GMT

Encountering very long processing times when trying to assess the proportional hazards assumption in a Cox regression model. I am having to terminate the process due to long wait times. The problem persists regardless of whether I am using SAS Base 9.4 (on Windows desktop) or SAS Studio (Mac). Below is the code, where &cat is a list of categorical variables (n=23), and &cont is a list of continuous variables (n=3).

proc phreg data=work.outcomes_fnl3_wocong;

class
demo_age1 (REF='<50') demo_gender (REF='M') primdx (REF='Ischemic')
preop_cr1 (REF='<0.9') preop_bili1 (REF='<1.2')
comorb_dm (REF="0") comorb_dial (REF="0") comorb_copd (REF="0") comorb_htn (REF="0") comorb_cvd (REF="0")
preop_abo (REF='A') preop_pra_sens (REF="<20") 
preop_ino (REF="0") preop_ecmo (REF="0") preop_iabp (REF="0") preop_vent (REF="0") 
cvs_prev_typ (REF="VALVE")
donor_age1 (REF='<50') donor_gender (REF='M') donor_cod (REF='Head Trauma')
induct (REF='0')
postop_plasma (REF='0')
op_era (REF="2004-2010")
 / param=ref;

model mort_y*mort(0)=&cat &cont / rl SELECTION = STEPWISE SLS = .05 SLE = .05;

assess var=(&cont)ph / resample;

run;

Re: Long lag time to assess proportional hazards assumption

ChrisNZ — Thu, 04 Mar 2021 19:52:30 GMT

Re-titled, and moved to procedures community.

Re: proc PHREG takes too long to assess proportional hazards assumption

Rick_SAS — Thu, 04 Mar 2021 20:01:20 GMT

How long does it take to run if you just fit one model instead of using stepwise selection?

Re: proc PHREG takes too long to assess proportional hazards assumption

SteveDenham — Fri, 05 Mar 2021 13:14:08 GMT

In particular, how long to fit the full model? I suspect that you may have to make some subject matter decisions regarding the independent variables to get this down to a workable size.

SteveDenham