Nest survival, model with "nesting season" and "year"?

[Florent]

Hello,

I rolled several models but I do not know how to build a model that would look at the nesting season of the year. In Chapter 17 of the MARK book, "nest survival" model 5 (page 17-4 MARK in the book) shows us how to estimate survival during the nesting season. To do this we must open our window PIM and ensure that the values ​​are all different (from 1 to 86 for our case). Then we will create a "Reduce design matrix" with two columns and 86 rows, the first being supplemented with "1" and the second being completed with numbers running consecutively from "1 to 86" in our case. Then you roll the model. If we did another analysis looking only effect of the year we will have a constant PIM (only "1" and not numbers of "1-86") and yet here we will use a "design matrix "which this time will include a single row with a column with the number" 1 "and another column with the name of my covariate" Year ".

The problem is that I don't know how to build a model that includes the nesting season (design matrix with 2 columns and 86 rows) and the covariate "Year" (design matrix consisting of a single row and 2 columns). I wonder if it would make the model and fill the colum "Year" on line 86??

Thank you in advance for your help.

Sincerely,

Florent.

[birdman]

It sounds as if you need to read the first 6-8 chapters of the Gentle Introduction document to get familiar with building additive models in the design matrix (DM).

In this case, retrieve your "season" model with the intercept (column of 1s) and the season effect (column of continuous 1-86). This is basically a linear trend (T) model as discussed in the manual, representing an incremental change in survival each day of the season. Add a third column to the original two, and call your "year" covariate in each cell of the column. This is an additive model of the seasonal trend plus year. An interactive model would be even more complex, and suggests that the seasonal trend differs across years.

As an aside, the seasonal trend model might be a good fit... they often are, but they use a lot of parameters to see the trend. Does daily survival change in a biologically meaningful way each day, or could you develop more parsimonious models of weekly, bi-weekly, or even monthly incremental differences in survival? Just an observation. If building such models in the DM seems daunting, you definitely have some reading to do in the early part of "the book".