Re: half-logistic distribution

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Re: half-logistic distribution

Therneau, Terry M., Ph.D.


On 12/26/2014 05:00 AM, [hidden email] wrote:
> i want to analyse survival data using typeI HALF LOGISTIC
> DISTRIBUTION.how can i go about it?it installed one on R in the
> survival package didn't include the distribution...or i need a code to
> use maximum likelihood to estimate the parameter in survival
> analysis.a typical example of distribution other than that installed
> in R will help.thanks
>

I am the author of the survival package, and had never heard of the "type I half-logistic"
before.
New distributions can be added to survreg() if they can be represented as location-scale
families; I don't think that your distribution can be written in that way.

Look at "survival" under the "Task Views" tab (upper left corner) of the cran.org web page
-- someone else may have already done it.

Terry T.

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
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.
Reply | Threaded
Open this post in threaded view
|

Re: half-logistic distribution

bbolker
Therneau, Terry M., Ph.D. <therneau <at> mayo.edu> writes:

>
>
> On 12/26/2014 05:00 AM, r-help-request <at> r-project.org wrote:
> > i want to analyse survival data using typeI HALF LOGISTIC
> > DISTRIBUTION.how can i go about it?it installed one on R in the
> > survival package didn't include the distribution...or i need a code to
> > use maximum likelihood to estimate the parameter in survival
> > analysis.a typical example of distribution other than that installed
> > in R will help.thanks
> >
 
> I am the author of the survival package, and had never heard of the
> "type I half-logistic" before.  New distributions can be added to
> survreg() if they can be represented as location-scale families; I
> don't think that your distribution can be written in that way.  Look
> at "survival" under the "Task Views" tab (upper left corner) of the
> cran.org web page

  I don't know about 'type I' but based on

http://en.wikipedia.org/wiki/Half-logistic_distribution

   you could try

dhalflogist <- function(x,s,log=FALSE) {
    r <- log(2)-log(s)-x/s-2*log(1+exp(-x/s))
    if (log) r else exp(r)
}
phalflogist <- function(q,s) { (1-exp(-q/s))/(1+exp(-(q/s))) }
rhalflogist <- function(n,s) { abs(rlogis(n,scale=s)) }
set.seed(101)
rr <- rhalflogist(10000,2)
hist(rr,freq=FALSE,breaks=80,col="gray")
curve(dhalflogist(x,2),add=TRUE,col=2,lwd=2)

d <- data.frame(rr)
library("bbmle")
m1 <- mle2(rr~dhalflogist(s=exp(logs)),start=list(logs=0),data=d)

  If you want to fit multiple groups etc., see the 'parameters'
argument of ?mle2

______________________________________________
[hidden email] mailing list -- To UNSUBSCRIBE and more, see
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.