I am currently working on a cause specific hazard regressions (cox) model and my model currently violates the proportional hazards assumption for my primary independent variable.  I am trying to model the cox with a different distribution (such as square root or log) to see if I can find a way to improve my model's fit. I am currently using the proc phreg function to model this proc phreg data=work.hypertension covs(aggregate); class case (ref='0') incquint (ref='1') ruralnum (ref='0'); model timefromhypertension*AdEvent(0,2)= case incquint ruralnum/RL ties=efron; id case_num; run; Is there any way in SAS for me to model the cox and use a different distribution in order to improve its fit?   Thank you ... View more

I am a beginner SAS user and would really appreciate some help.   This is an example of the dataset I have for several thousands of patients: Patient_ID HF HF_date MI MI_date Stroke Stroke_date 1 0 . 1 31DEC2019 1 06AUG2005 HF=Heart Failure, MI=Myocardial Infarction   I am working on a survival analysis and will need to do it for a composite outcome called MACE (which will the 3 variables in my dataset: HF, MI and stroke). I am trying to find the best way to create a MACE_date that reflects the earliest date if a patient has more than 2 outcomes. So for patient 1 above, I would want the MACE_date to be 06AUG2005 instead of 31DEC2019.   I currently have: MACE=0; if HF=1 then MACE=1; if MI=1 then MACE=1; if stroke=1 then MACE=1;   MACE_date=.; if HF=1 then MACE_date=HF_date; if MI=1 then MACE_date=MI_date; if Stroke=1 then MACE_date=Stroke_date;   How can I code my SAS in a way where if a patient has multiple outcomes (HF and MI and stroke), that the MACE_date will reflect the earliest date? Is there an easy way to do this?   Thank you so much in advance ... View more