DSR varies across the nesting season

[janjedlikowski]

I'd like to ask you, what is the best method to estimate nest success from the nest survival model in which DSR varies across the nesting season (linear or quadratic time trend)? Take mean DSR and raise it to the power (to the length of the period of concern), or maybe find mean nest initiation date and calculated the probability that a nest would survive the entire incubation period from that date, or maybe there are other better ways?

Thank´s
(sorry for my terrible English)
Jan

[gstauffer]

It probably depends somewhat on the strength of time trends. I'm sure others have tried various methods, but see these 2 papers for examples.
Dinsmore et al. (2002. Ecology 83:3476–3488) calculated nest success (NS) from the mean initiation date. However, because the probability of finding the nest changes throughout the season if DSR changes, they used a Horvitz-Thompson estimator to adjust the mean initiation data.
Stauffer et al. (2011. J Wild Manage 3: 548–557) also used an H-T estimator to predict nest initiation dates for nest that were not found, but then calculated a weighted mean NS from all nests based on the estimated distribution of nest initiation dates.
At the time I did some simulations to compare these two methods for a few different scenarios (low/high DSR, many/few nest found, weak/stronger temporal trends). In general, the second method seemed to perform better, especially when there were strong quadratic trends (higher-lower-higher) in DSR, in which case use of the H-T-adjusted mean initiation date did not improve estimates estimates based on the naïve mean nest initiation date.
Another alternative that worked pretty well when trends were only slight to moderate was to simply use the constant Mayfield estimator (or a comparable constant DSR model).

[bacollier]

Jan,
There are several ways, so you have options, here is one I did for hen survival analysis using the nest survival model approach that has a continuous covariate that is different for each individual in the sample set. You might also look for work by B. Sandercock (Kansas State) as he has a few papers that use this type of approach as well.

Collier, B. A., K. B. Melton, J. B. Hardin, N. J. Silvy and M. J. Peterson. 2009. Impact of reproductive effort on survival of Rio Grande wild turkey hens in Texas. Wildlife Biology 15:370-379

I worked up this code snippet (in R here, note it requires package rgl) that creates roughly the graph that I never got to put in the paper. I hope it helps.

Code: Select all

install.packages("rgl")
library(rgl)
for(Days_Nesting in 1:55){
   Breeding_Period<-1:118
   DSR_RGWT= exp(6.6205449-0.0589883*Days_Nesting)/(1+
            exp(6.6205449-0.0589883*Days_Nesting))
   DSR_BS<-DSR_RGWT^Breeding_Period
   DSR_Final<-as.data.frame(cbind(Days_Nesting, Breeding_Period, DSR_BS))
   write(c(t(DSR_Final)), "location.txt", append=T, ncolumns=3)
}
dsr_data<-read.delim("location.txt", header=F, sep="")
attach(dsr_data)
head(dsr_data)
plot3d(V1, V2, V3)


\bret

[janjedlikowski]

Thank you for reply.