Dear list, I have data on insect survival in different cages; these have the following structure: deathtime status id cage S F G L S 1.5 1 1 C1 8 2 1 1 1 1.5 1 2 C1 8 2 1 1 1 11.5 1 3 C1 8 2 1 1 1 11.5 1 4 C1 8 2 1 1 1 There are 81 cages and each 20 individuals whose survival was followed over time. The columns S,F,G,L and S are experimentally manipulated factors thought to have an influence on survival. Using survfit(Surv(deathtime,status)~cage) gives me the survivorship curves for every cage. But what I??d like to have is a mean survivorship value for every cage. Obviously, using tapply (deathtime,cage,mean) gives me mean values, but I??d like to have a better estimate of this using a proper statistical model. I??ve tried a glm with poisson errors (as suggested in Crawley??s book, page 628), but the back-transformed estimates (using status as the response variable and deathtime as an offset) were totally unrealistic. As I??m new to survival analysis, it would be great if anyone could give me some hints on what method would be best. Thanks a lot! Christoph
On 11/17/05, Christoph Scherber <Christoph.Scherber at uni-jena.de> wrote:> Dear list, > > I have data on insect survival in different cages; these have the > following structure: > > deathtime status id cage S F G L S > 1.5 1 1 C1 8 2 1 1 1 > 1.5 1 2 C1 8 2 1 1 1 > 11.5 1 3 C1 8 2 1 1 1 > 11.5 1 4 C1 8 2 1 1 1 > > There are 81 cages and each 20 individuals whose survival was followed > over time. The columns S,F,G,L and S are experimentally manipulated > factors thought to have an influence on survival. > > Using survfit(Surv(deathtime,status)~cage) gives me the survivorship > curves for every cage. But what I伮伌d like to have is a mean survivorship > value for every cage. > > Obviously, using tapply (deathtime,cage,mean) gives me mean values, but > I伮伌d like to have a better estimate of this using a proper statistical > model. I伮伌ve tried a glm with poisson errors (as suggested in Crawley伮伌s > book, page 628), but the back-transformed estimates (using status as the > response variable and deathtime as an offset) were totally unrealistic. > > As I伮伌m new to survival analysis, it would be great if anyone could give > me some hints on what method would be best.No method is best, but some methods may be useful ;) One such may be to fit a parametric model to your data. Check 'survreg'. G伱伓ran> > Thanks a lot! > Christoph > > ______________________________________________ > R-help at stat.math.ethz.ch mailing list > stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide! R-project.org/posting-guide.html >-- G伱伓ran Brostr伱伓m