Estimating phi (phi = S*psi) in a multi-state model in MARK

questions concerning analysis/theory using program MARK

Estimating phi (phi = S*psi) in a multi-state model in MARK

Postby jbauder » Wed Feb 07, 2018 1:45 pm

Hello,

I am trying to replicate and expand a multi-state survival analysis where the researchers used the phi parameterization of the multi-state model in SURVIVE to estimate phi (where phi = S*psi). They used a radio telemetry data set and a multi-state model where the states where 1=located alive, 2=located but status unknown (e.g., in a burrow), and 3 = dead. They report survival as the probability of being alive at time i and staying alive until time i + 1 (phi alive-alive). Apparently there were a sufficient number of telemetry locations in the unknown state so that a known-fate model was not appropriate. I have some new data to add to the original analysis but as a MARK/RMark user who has never used SURVIVE, I first wanted to look into doing this analysis in MARK. I found this post from 2009 (viewtopic.php?f=1&t=1265&p=3549&hilit=+phi+multi+state+#p3549) where Evan explained that MARK does not provide an estimate of phi (although he mentioned it could be possible to calculate phi by hand). I wanted to follow up on this post to see if there have been any new developments or if anyone has any additional suggestions on how one might conduct this analysis in MARK before trying to learn SURVIVE.

Thanks,
Javan

P.S. Here is the citation for the previous study if anyone is interested:
Breininger, D. R., M. J. Mazerolle, M. R. Bolt, M. L. Legare, J. H. Drese, and J. E. Hines. 2012. Habitat fragmentation effects on annual survival of the federally protected eastern indigo snake. Animal Conservation 15:361-368.
jbauder
 
Posts: 56
Joined: Wed May 25, 2011 12:01 pm

Re: Estimating phi (phi = S*psi) in a multi-state model in M

Postby jbauder » Wed Feb 07, 2018 2:04 pm

And I am very sorry for accidentally posting a (near) duplicate of this post earlier!!! :( Please forgive my hastiness, and please ignore the earlier post!
jbauder
 
Posts: 56
Joined: Wed May 25, 2011 12:01 pm

Re: Estimating phi (phi = S*psi) in a multi-state model in M

Postby ganghis » Wed Feb 07, 2018 2:20 pm

Hi Javan,
It looks like Jim Hines coded it up in a special version of SURVIV (similar to MSSURVIV). I'd contact him directly if you want to try to get your hands on the exact code he used, which will be a lot more user friendly than SURVIV. You DO NOT WANT to use SURVIV directly. Believe me.

You could also probably implement the exact model from the paper in ESURGE, where you have the option of estimating phi or decomposing it into phi = S * Psi.

Hope this helps a little.

Cheers, Paul
ganghis
 
Posts: 84
Joined: Tue Aug 10, 2004 2:05 pm

Re: Estimating phi (phi = S*psi) in a multi-state model in M

Postby darryl » Wed Feb 07, 2018 3:33 pm

You DO NOT WANT to use SURVIV directly. Believe me.


Come on, Paul. Everyone should have to use SURVIV at least a couple of times to realise how far we've come. :wink:
darryl
 
Posts: 495
Joined: Thu Jun 12, 2003 3:04 pm
Location: Dunedin, New Zealand

Re: Estimating phi (phi = S*psi) in a multi-state model in M

Postby jhines » Wed Feb 07, 2018 5:27 pm

Hi Javen,

I saw the phi.dot post and was preparing to respond, but your email came before I was able to send :)

I don't have a copy of the final data that Marc used in the paper, but I do have the program. Program SURVIV was the pre-cursor to program MARK, meaning it could analyze many different types of user-specified models, but has the drawback that specifying a model could be very cumbersome. I created a modified version specifically designed to handle multi-state models, making the model specification a little easier, but still not as nice as MARK. For the indigo snake analysis, I modified the multi-state version a little more to make the input parameter specification similar to MARK (using PIMs). If your data is similar to the data in the paper, then it might be easiest to use that software. The main problem is that I did not write the program for general use, so there is no documentation. I think I could write some instructions for it if you decide to use it.

The other option is to use program ESURGE (as mentioned by P. Conn). This program was specifically designed for multi-state analyses and would be my choice if I had to do it over again. There is a learning curve in using this software, but once you get the hang of it, model specification is very easy. One drawback of this software is that it was written in Matlab. This means you need to download a very large Matlab library in addition to the ESURGE software. However, with a reasonably current PC or laptop and decent internet connection, this isn't a major problem. It also means that the program will only run on a Windows PC. If you only have a Mac or Linux machine, you cannot use it (unless you can dual-boot the Mac or run a Windows emulator).

One other note: I developed an R package (R2Esurge) to interface with ESURGE for a recent workshop. I don't think anyone has tried it yet, but it is available if you're an R user.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Estimating phi (phi = S*psi) in a multi-state model in M

Postby jbauder » Wed Feb 07, 2018 6:14 pm

Hi Dr. Hines,

Thanks for replying and for the update on SURVIV and ESURGE. I am glad to hear that there is an R package that uses ESURGE.

Thanks,
Javan
jbauder
 
Posts: 56
Joined: Wed May 25, 2011 12:01 pm


Return to analysis help

Who is online

Users browsing this forum: No registered users and 12 guests