Pradel model question

[cnicolai]

I've been trying to partition sources of recruitment from different groups in my analysis. I've got 9 groups in which 8 are individuals from known sources and a last group which is from unknown sources. I've identified the potentially correct parameterization for the phi and p parameters and I need to deal with the recruitment parameter. In a model in which I build no group variation, I get a certain estimate of recruitment. Then, when I build a model allowing for group variaiton in recruitment, I get 9 group-specific estimates. The big question is, how can I partition these sources of recruitment? I assume they are not additive, because when I sum the 9 groups, the estimate of recruitment is way higher than from the no group model. Is using the product of the group estimates the way to get overall recruitment as in the non-group model?

Chris

[gwhite]

Chris:
What you are trying to do is more complicated than just a simple function of the f estimates (I'm assuming you are using the f, phi, p parameterization of the Pradel model. Study closely the definition of the f estimates: new animals per old animal. Nichols always defines it as new guys per old guy. So f is the ratio of new animals per old animal in the population. Therefore, the f estimates are not additive. To partition it, you need to compute the number of new animals (B) for each occasion for each model, and then sum the B values. The following document will help you understand these definitions better:

http://www.cnr.colostate.edu/class_info ... pradel.pdf

Also, check the MARK help file.


Gary

[cooch]

Chris:
What you are trying to do is more complicated than just a simple function of the f estimates (I'm assuming you are using the f, phi, p parameterization of the Pradel model. Study closely the definition of the f estimates: new animals per old animal. Nichols always defines it as new guys per old guy. So f is the ratio of new animals per old animal in the population.

In fact, this is noted explicitly in Chapter 13 - p. 3:

"Now, lambda and phi are familiar (and explicitly defined
above). What about B(i)/N(i)? This is the per capita rate of
additions to the population (often referred to somewhat 'sloppily'
as the recruitment rate, which has a very specific demographic
meaning that is often ignored-for purposes of consistency with some
of the literature, we'll ignore it too). It is the number of
individuals entering the population between (i) and
(i+1) (i.e., B(i)) per individual already in the population
at time (i) (i.e., N(i)). Let's call this recruitment rate
f(i)..."

Therefore, the f estimates are not additive. To partition it, you need to compute the number of new animals (B) for each occasion for each model, and then sum the B values. The following document will help you understand these definitions better:

http://www.cnr.colostate.edu/class_info ... pradel.pdf

Also, check the MARK help file.

...and Chapter 13, where the 'algebra' is worked out, and all the parameters defined.