Robust - RMARK diff estimate of survival from MARK

[amandarg]

Sorry, hopefully this is the last problem I will be having with this model. I ran all my models and noticed that my estimates of apparent survival (which is what I am ultimately interested in) were sometimes very different then when I was putting them into MARK by hand using design matrices. I did this because one of my estimates in RMARK (the last one for S) seemed way too high based on my capture history but it about what I expected in MARK. The two design matrices are similar and I had Brett Sandercock look at them as well so make sure I didn't just make a mistake. Neither of us are quite sure what is going on and which estimates I should be using. I made copies of the DM from both programs that I can send that should be modeling the same thing. I'm also attaching the code I ran just for one of the models that I checked. I wanted to run the model: S(time) pi(session) m(bh) t(session)

Code: Select all

#RMark analysis
#Load RMark package
library(RMark);
rm(list=ls())
cleanup(ask=FALSE)

#inport .inp
bentsadults=convert.inp("C:/R_robust/inpfiles/bentsfixed_final.inp",
group.df=data.frame(sex=c("Male", "Female")))

#Process the data so models can be analyzed
#Time intervals for 6 primary occasions (300+ collectors)
#use the format from MARK where you put in the x/60 for time intervals
time=c(0,0,0,0,0,0,.733,0,0,0,0,4.25,0,0,0,0,0,0,0,0,0,0,0,0,1.47,0,0,0,0,0,0,0,0,0,0,0,0,0,4.47,0,0,0,0,0,0,0,0,0,0,0,0,.9,0,0,0,0,0,0,0,0,0,0,0,0,0)
bentsadults.proc=process.data(bentsadults,model="RDFullHet",time.intervals=time,
groups="sex")
summary(bentsadults.proc)

bentsadults.robust.run=function()
{

#Create the design data
bentsadults.ddl=make.design.data(bentsadults.proc)

#Create parameter specifications for formulas

S.time=list(formula=~time)

# "No Emigration" - set Gamma'' = Gamma' = zero
GammaDoublePrime.GammaPrime.nomove=list(GammaDoublePrime=list(formula=~1,fixed=0,share=TRUE))
GammaDoublePrime.GammaPrime.nomove=list(GammaDoublePrime=list(formula=~1,fixed=0), GammaPrime=list(formula=~1,fixed=0))

#m(bh) diff each primary period
p.mixturesession=list(formula=~session:mixture+c,share=TRUE)

pi.session=list(formula=~session)


N.session=list(formula=~session)

# create model list
cml=create.model.list("RDFullHet")
#row.names(cml)=c(1:length(cml$p))

# run and return models
results=mark.wrapper(cml,data=bentsadults.proc,ddl=bentsadults.ddl,adjust=TRUE)
return(results)
}

robust.results=bentsadults.robust.run()
robust.results

results.table=robust.results$model.table

export.MARK(bentsadults.proc, project.name="bentsrobustf", robust.results,
replace=TRUE)







[jlaake]

I'm not entirely sure what you were saying in your message other than that you are getting differences between what you got in MARK and RMark. There are at least 4 reasons for differences assuming you did everything in MARK correctly:
1) different data
2) different model
3) different link functions
4) different DM for the same model

The last 2 typicallu only make small differences. From looking at what you show, my guess would be different model. You have used groups="sex" but then your model for N only includes session. To me that seems strange that you wouldn't expect differences in abundance across sex. Note that the model for N is actually for f0 so even if abundance is the same, f0 can differ if one sex is more likely to be caught than the other. I would start with looking for differences in the models that you created with MARK and RMark.