BookmarkSubscribeRSS Feed
CHELS
Obsidian | Level 7

Suppose I have time to event data for 100 subjects, with event indicator (1=Event, 0=Censored) and a Time variable (maximum follow-up of 2 years). 

I also have a risk score for each of the subject and  I'd like to generate time-dependent ROC and get the AUC at time = 2 years.

However, all of the events happened within the first year and no events between 1-2. With the below code, SAS is only generating AUC up to 1 year. 

 

proc phreg data=Indat concordance plots=roc rocoptions(at=2) ;
model TIME*EVENT(0)=HM3RS;
run;

 

It gives a warning that say "you have chosen At= time greater than the maximum observed. " Can someone help me to understand this? I thought for Cumulative/Dynamic ROC, everything before 2 years can be taken into account, even though there is no event between 1-2 year. 

 

Thank you!

1 REPLY 1
sbxkoenk
SAS Super FREQ

Values for AT= option should be between minimum and maximum observed time!

 

Paper SAS462-2017
Evaluating Predictive Accuracy of Survival Models with PROC PHREG
Changbin Guo, Ying So, and Woosung Jang, SAS Institute Inc.

https://support.sas.com/resources/papers/proceedings17/SAS0462-2017.pdf

 

BR, Koen

hackathon24-white-horiz.png

The 2025 SAS Hackathon has begun!

It's finally time to hack! Remember to visit the SAS Hacker's Hub regularly for news and updates.

Latest Updates

What is ANOVA?

ANOVA, or Analysis Of Variance, is used to compare the averages or means of two or more populations to better understand how they differ. Watch this tutorial for more.

Find more tutorials on the SAS Users YouTube channel.

Discussion stats
  • 1 reply
  • 584 views
  • 1 like
  • 2 in conversation