plotting estimates as function of environmental covariates

[cooch]

I just posted a new version of Chapter 6, which covers the mechanics of how you would plot the value of a parameter against a particular environmental covariate that might be of some interest. While the capability plotting parameter values against individual covariates has been in MARK for some time (and covered in some detail in Chapter 11), there hasn't been much said in 'the book' about what to do if you wanted to plot the parameter as a function of a group (environmental) covariate.

The new Chapter 6 corrects this omission. Section 6.8.2 runs through everything for a single model. You can either do this 'by hand', or, by treating environmental covariates as individual covariates, and using the individual covariates capabilities in MARK.

Perhaps the more important addition (which builds off material in 6.8.2) is new section 6.16, which takes the same idea, but now consider model averaging. This is a fairly big deal, since there have been a fair number of questions over the years on whether or not you should try to model average beta's for a particular covariate of interest (general conclusion -- don't. See viewtopic.php?f=1&t=996 for a fairly complete discussion of some of the issues). So, if you focus on the real parameter estimates, how do you generate a plot of model averaged real estimates (on the vertical axis) against the continuous, environmental covariate (on the horizontal axis)? Actually, the model averaged estimates themselves are trivial to calculate. What is far less trivial (well, its easy, but laborious) is estimated the SE and 95%CI for these model averaged estimates. Section 6.16 goes through all of it in some detail. Again, doing it by hand (as presented in 6.16) isn't hard, but laborious, especially if you have a large candidate model set.

At some point, I'll write up how you can 'speed things up considerably, by using the 'environmental covariates as individual covariates' trick. Much faster, although you do need to re-format your data.

At any rate, if you've wondered about how to 'describe' the relationship between some parameter and an individual covariate, the new bits in Chapter 6 should help.