I'm trying to build up a survival curve using Cox PH Regression. However, there are no categorical independent variables (no groups to compare such as gender or treatment for example) in my model. Only numeric independent variables such as age, and temperature. Can I still use Cox PH Regression for this purpose? Or should I Weibull or Kaplan-Meir instead?
You can run cox regression. Just check the assumptions. David Booth
once you have properly coded your "Group" categorical variable, so that it only has the four values you mentioned, you can run a Cox Regression analysis in SPSS specifying that the variable is to be treated as categorical. In order to do this, click on the "Categorical..." button in the analysis main panel, then move the Group variable to the right (Categorical Covariates), select "First" as reference category, confirm this selection pressing "Change", and finally press ok and run the analysis. This way, it should output three different hazard ratios, for every group other than the first.
Emanuele Cassioli
No categories were included in the analysis. If you run the regression in SPSS without defining any category (only numerical values), you can still come up with a survival function.
The hazard ratios have no interpretation for categories in this case! I was wondering if that is a valid approach!
For example, my Event diffinition was 1 if a pavement section has shown a specific malfunction within a certain time, and 0 if censored. My independent variable was the thickness if the section as X1 only. No treatment was applied for comparison. I just have X1 as a thickness only which is a numeric value.
Cox regression estimates harazd ratio (HR), not cumulative survival.
Kaplan-Meier , Nelson-Aalen estimators calculate approxmatly cumulative survival function ( life table does the same).
So... as You have no pedictors , Cox regression has nothing to do with it ....
Bonus: Cox regression assumption about proportionality could be violated while You use Kaplan-Meier and others estimators :)
Thank you for your answer. I have predictors, but they are continous variables, not categorical, and there is also a way to build a survival curve with cox regression.
https://m.youtube.com/watch?v=U3hiHALQjAI&t=28s
of course, we can predict survival curve based on values of variables ( new patient for example). In the video survival curve plotted for mean of continuous variables.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression