[R] Survreg/psm output

Frank E Harrell Jr f.harrell at vanderbilt.edu
Thu Apr 16 21:39:40 CEST 2009


Laura Bonnett wrote:
> Dear R-listers,
> 
> I know that there have been many, many posts on the output from
> Survreg.  To summarise what I have read, Scale is 1/shape of the
> Weibull which is also the standard deviation of the normal
> distribution which is also the standard deviation of the log survival
> time and Intercept is log(scale).  I also know that the hazard
> function can be calculated from the output to give something like:
> h(time)=exp(-intercept)^scale x scale x time^(1-scale)
> So, for example if the intercept was -1.11 and the scale was 1.17 then
> the hazard function is h(t) = exp(1.11)^1.17 x 1.17 x t^0.17 = 4.30
> t^0.17
> 
> However, how can I work out what the hazard is I.e. by how much does
> the risk of death increase or decrease with time? Some people have
> said that is the scale parameter is greater than 1 then the risk of
> death increases with age.  Is this correct?
> 
> I am also aware that Harrell has posted previously that there is a
> case study in his book.  I have a copy of the book and have read all
> the case studies but I'm still not convinced I know what the hazard of
> death is?!

library(Design)
f <- psm(...)
h <- Hazard(f)  # fitted hazard function as an R function
h    # look at it
Evaluate h at any time and any linear predictor (the latter being 
computed by for example predict(f, newdata)).

Frank

> 
> Sorry that this relates to previously posted queries, but as yet there
> doesn't seem to be a satisfactory solution to the question of the
> output?
> 
>  Thank you for your help as always,
> 
> Laura
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
> 


-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University




More information about the R-help mailing list