Running MIFitHMM with tracks from Animotum

[mitchrider]

Hello,

I'm wondering if there is a way to use MIFitHMM with tracks reconstructed by Animotum to account for the measurement error of the predicted locations.

Thank you!
Mitch

[bmcclintock]

Yes, in theory MIfitHMM will accept any list composed of momentuHMMData objects returned by prepData. So you'd just need to run prepData on each simulated track from {aniMotum} and then combine them into a list for MIfitHMM. I've not done this myself yet, but I was hoping to eventually include an example demonstrating exactly this in the {momentuHMM} vignette. So please let me know if you have any difficulties!

[mitchrider]

Thanks for getting back to me! I was able to do this with simulations from the sim_fit function in animotum. However, I realize those simulations don't constrain the tracks like crwSimulator and crwPostIS so I'm assuming I need to figure out a way to format that animotum output so I can use those functions to create simulations? Have you been able to do this?

[bmcclintock]

Are you just fitting the (discrete-time approximation of the) continuous-time correlated random walk (CTCRW) model in {aniMotum}? If so, maybe you should just use {crawl} to fit the CTCRW (crawl::crwMLE) and simulate from it (crawl::crwSimulator --> crawl::crwPostIS). Note that momentuHMM::crawlWrap is just a wrapper for these three {crawl} functions. Otherwise, have you looked into the {ctmm} package? I haven't looked into too much yet, but I think you can fit the CTCRW ("OUF" in {ctmm} parlance) and draw realizations of the observed path via simulation.