Multi-state models with age structure

Forum for discussion of general questions related to study design and/or analysis of existing data - software neutral.

Multi-state models with age structure

Postby X33277 » Tue Mar 12, 2019 9:32 pm

Hello, I am a new Mark user and inquiring about how to best set up a multi-state model with age structure. I have gotten as far as chapter 13 in the Mark book over the past several months but obviously have a long way to go..... In this current study there are 3 states with 2 age classes. I thought maybe about 6 states with each age class having its own 3 states and constraining psi so that the older age class cannot transition to any state of the younger age class. Perhaps 3 states with 2 age groups would work as well but not sure if it would be more or less appropriate. I am sure the best approach is obvious to most here.... so I am hoping for some suggestions.
Thanks for any advice.
Posts: 2
Joined: Tue Mar 12, 2019 9:02 pm

Re: Multi-state models with age structure

Postby simone77 » Tue Mar 19, 2019 5:48 am

I am afraid you have not been clear enough as to make possible someone can help you. Think of a state as some individual trait that changes in a dynamic and stochastic (non-deterministic) way along its life. Age is a dynamic individual trait but it is not a state (in capture-recapture jargon) because it changes in a deterministic way and, therefore, the model is not going to estimate the probability of transition between, say, age 1 to age 2.

A group, instead, is some fixed individual trait (it does not ever change as, for instance, sex in many species).

Age, group and states are treated in a different way in MARK (or any other software for capture-recapture modelling). States sometimes are related to age when, for instance, your state refer to the breeding status (non-breeder and breeder) but in many cases they don't. Also, six states are quite a few and it may mean a lot of parameters to be estimated (p from s1 to s2, p from s2 to s1, p from s1 to s3, ..., p from s6 to s5), is that realistic with the data you handle?

Anyway, it is really difficult to help you if you do not explain in mode detail your study system and your goals. May you please try?
Posts: 153
Joined: Mon Aug 10, 2009 2:52 pm

Re: Multi-state models with age structure

Postby X33277 » Fri Apr 26, 2019 7:18 pm

Hi Simone,
Thanks for the reply and the information. To give some additional details there are 3 biologically meaningful states for the species but with 2 distinct age classes. Survival is usually constant over time but differs between age classes with prime age adults ages 5-19 years of age having higher survival than individuals 20+. We know the exact age of all the individuals. The primary reference I am using here is Chapter 7 in the mark book. During the sampling any aged individual is captured so TSM doesn't work. The example in chapter 7 where individuals are marked as both young and adults are treated as 2 groups. But since in this example group 1 (ages 5-19) will eventually move into group 2 (Ages 20+) I am wondering if that negates treating the age classes as groups. Not sure if using age at capture as an individual covariate would work? I have not used RMark before but perhaps that is what I need to do and create age bins? Doesn't seem like something I can build with the PIM's.
Thanks for any insight.
Posts: 2
Joined: Tue Mar 12, 2019 9:02 pm

Re: Multi-state models with age structure

Postby jlaake » Fri Apr 26, 2019 8:13 pm

You can use a time varying individual covariate for age if you want. What you do is create an individual covariate for each occasion and fill in the values of the age of each individual at each occasion. The individual covariates are put in the data. Values before the first capture are not used so you can use 0 or any other numeric value. If you use MARK interface, you put the covariate name in the DM in the row associated with that occasion/ interval. If you use RMark, then the covariate names must have a common prefix(eg iage) and the suffix must match the time value so RMark knows which covariates to put in which row of the DM. In the RMark formula you use the prefix (eg iage) in the formula.
Posts: 1088
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Return to analysis & design questions

Who is online

Users browsing this forum: No registered users and 1 guest