Can anyone tell me if it's possible to fit a multi-state robust design model in MARK allowing for state transitions to occur between secondary occasions within primary periods?
Here's the setup... My post-doc and I are working with a long-term data set in which surveys have been conducted every summer for about 15 years. Approximately 20 to 30 surveys have been conducted each year, but they are weather dependent so the time between varies. Individuals are long-lived and we assume that the population is closed over each summer. However, the student area includes several different locations and individuals may move between locations between surveys within the same year.
We'd like to fit a model that is essentially a MSCRD but allows for the movement between states within a single primary period. The MSCRD in MARK doesn't allow this, so we've fudged it with a simple MS model by fixing the survival probabilities to be 1 between occasions within a secondary period.
However, there is a challenge handling the transitions between years. We need to allow the transition probabilities within years and the transition probabilities between years to differ. Specifically, we assume that the transition probabilities depend on the distance between the starting and ending locations within years when the time between surveys is a matter of days. However, the starting location should not have any effect on the transitions between years because there will be ~250 days between surveys.
Using an R(Mark) like formula, I'd like to construct a model of the transition probabilities like:
- Code: Select all
~ (1-last) * stratum:tostratum + last * tostratum
where last is dummy variable indicating if the current survey is the last in the year. This would allow the transition probabilities to depend on the starting and ending states within years, but only on the ending state between years.
The problem with doing this is that it requires using time dependent PIMS instead of constant PIMS for the transition probabilities. With 50-60 entries in the capture probabilities per year (we include dots to account for missed days), 15 years, and 15 locations the PIMS are huge!
Can anyone tell me if there is an easier way to do this? Is there an existing model that allows for transitions between the surveys within a single year that I'm missing?
Thanks in advance!
Simon