max occasions in MARK/RMARK

[ganghis]

Hi all,
A grad student and I are looking to use multistate models to explore fish movement using passive accoustic arrays. The number of fish is relatively low (40), but we were thinking of using a lot of sampling occasions (e.g. 3000-4000) to get at fine scale movements. However, it would appear that MARK doesn't like this (it won't let me set up simulations with numbers in the 100s). Obviously we'd have to go the RMark route to avoid dealing directly with PIMs and DMs but is there a hard upper bound on the number of occasions we should know about before constructing encounter histories that wrap around the sun?

Paul

[egc]

Hi all,
A grad student and I are looking to use multistate models to explore fish movement using passive accoustic arrays. The number of fish is relatively low (40), but we were thinking of using a lot of sampling occasions (e.g. 3000-4000) to get at fine scale movements. However, it would appear that MARK doesn't like this (it won't let me set up simulations with numbers in the 100s). Obviously we'd have to go the RMark route to avoid dealing directly with PIMs and DMs but is there a hard upper bound on the number of occasions we should know about before constructing encounter histories that wrap around the sun?

Paul

In theory, I could recompile MARK.EXE under Linux with as many occasions as you want. I'd need to work with Gary to figure out all the subroutines that would need changing. And, obviously, you'd need a Linux box to run it on.

[esnook]

I'm very interested to know what the solution to this problem was as I have the same one. Just wrote a post titled: "Encounter History Length with acoustic telemetry data".

Any ideas would certainly be appreciated!

Thanks!
Erin

[Fish_Boy]

We have been doing a lot of acoustic work lately as well. I was thinking that a sensitivity trial of sorts on different time intervals could be done to determine an 'optimal' time interval for the resolution of your data. We have run into this issue, for any number of experiments and studies, where determining the best time interval or spatial resolution to build deterministic models is of interest. Perhaps trying some longer time periods and gradually increasing them e.g., annual, monthly, daily, hourly. You could look at whatever parameters are of interest on an hourly basis for a subset of days monitored to assess the shorter time scales... that would give you an idea of the how sensitive the parameters are to time intervals, at least as an initial step.

[dhpstroud]

Hi, I'm interested to know what you end up doing. I am also using passive acoustic telemetry data and currently trying to determine my time scale for my multi-state analysis. The problem I keep running into pertains more to receiver detection overlap when I increase my time scale (from hours to days, for example). How have others dealt with this?

[rasrage]

Hi all,
I realize this is an old thread but I seem to be missing the final answer to it. Is there a maximum number of occasions MARK can handle, or a maximum length of the encounter history (including covariates, comments, etc) when accessed through RMARK? Similar to the previous posts in this thread I am trying to figure out what can realistically be done, in terms of CR analyses, with a very large robust design data set.
Thank you!
Rahel

[rasrage]

... to (partially) answer my own question - MARK (through RMark) seems to handle 10k individuals and 1300 occasions without problems, and only takes a few hours (for the simplest RD model). Impressive!