Incorp. seasonal T in real estimates of nest surv

[birdman]

I am running a nest survival analysis with a reasonably large dataset that is highly variable and supports fairly large, complex models. Based on AICc results and beta coefficients and confidence estimates, there are several potentially informative, overlapping temporal sources of variation. There is a weakly negative but well supported effect of nest age, a similarly supported effect of time since the last new moon (this is a tidally influenced system), and a negative seasonal trend (the T model in MARK lingo).
I’m now trying to run predictive models of survival under several scenarios to graphically depict our results. I have used the following code based on the example files to generate daily survival estimates. I simply set the time since the last new moon at the mean value in this instance, and use the seq.ages I created for a 25 day nesting cycle to create the design matrix (DM) for use in the compute.real call.

Code: Select all

for (iage in seq.ages) {
   fc$value[1:936]=iage
   colnum <- colnum+1
   fc$value[fc$var=="LastNew"]= 14   ### USER NEEDS TO SET TIME SINCE LAST NEW MOON OF INTEREST - MEAN = 13.95
   design <- fill.covariates(bestmod,fc)
   design[,10:14] <- design[,1]*design[,3:7]
   point[,colnum] <- compute.real(bestmod, design=design)[,"estimate"]
   lcl[,colnum] <- compute.real(bestmod, design=design)[,"lcl"]
   ucl[,colnum] <- compute.real(bestmod, design=design)[,"ucl"]
}

My question is about how to incorporate the seasonal trend. How do I set up the DM to estimate the above for a nest initiated on day 25 of the season, versus one set up much later, say day 75? The original DM in the estimation process runs from 1-116, the length of the season. It seems that to correctly incorporate the T effect, the DM has to include the daily incrementing, which occurs simultaneously with the aging of the nest. What am I missing? Thanks for any feedback you can provide.