Re: {MEDSTATS} Parametric vs Semi-parametric survival in prediction models

[Peter Flom]

Pedro Emmanuel Alvarenga Americano do Brasil  writes

 <<<

Im looking for a reading, as breif as possible, about how to decide wether using paramentric or semi-parametric survival models when the objective is to develop a prediction model. 

 >>>

From what I understand, semi-parametric models don't make predictions.  In fact, I think that was one of the main contributions of Cox' model - he realized you can get a good estimate of the hazard ratio without assuming anything about the underlying distributions except proportional hazards.  So the Cox model gives estimates of the hazard ratio, but not of survival time.

I am sure someone will correct me if the above is wrong

Peter

Peter L. Flom, PhD
Statistical Consultant
Website: http://www DOT statisticalanalysisconsulting DOT com/
Writing; http://www.associatedcontent.com/user/582880/peter_flom.html
Twitter:   @peterflom

[Jeff Allard]

Pedro- You can estimate survival probabilities from Cox regression using the baseline function.
I believe there are pros and cons of Cox versus parametric approaches. I recall Allison having a discussion in his book Survival analysis using the SAS system. I have decided between the two (and then there are many parametric forms to choose if you go that direction) based on fit validation and effectiveness for prediction, like any other regression model. My sense is that Cox is more flexible and has become the de-facto standard for many researchers.

2009/12/23 Peter Flom <peterflom...@mindspring.com>

--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .
MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules