RMark 1.7.4

[jlaake]

I've loaded RMark 1.7.4 on the website. Of particular note here is that the source (RMarkSource.zip) can now be downloaded and used on a Linux machine. The source can be found at the bottom of the RMark download page. At this point if you want to use RMark on a Linux machine you'll have to get mark.exe from Evan. Also, Linux users make sure you read the notes in Whatsnew. Let me know if you encounter any problems.

In addition to Linux support, a few minor fixes were included and a new function make.time.factor (v1.7.3) was created which was particularly useful with the weta example from the Occupancy book to create an observer factor variable that differed across sites/visits. Note that there is a discrepancy with one of the models between the weta results and the contents in the book. I spoke with Darryl and he knows about the error in Table 4.7.

As a side note, if you are using the MSOccupancy model be aware that Psi1 and Psi2 are defined differently by Nichols et al (2007) (also as in MARK) than what is described on the occupancy web site. This may cause confusion if you are trying to duplicate results with RMark/MARK and the exercise on the website. In Nichols et al (2007) Psi2 is a conditional probability such that the probability of being in state 2 is Psi1*Psi2 whereas in the occupancy web site example, Psi2 is the probability of being in state 2 which leads to the necessary constraint that Psi1+Psi2<=1. This constraint is not needed with the Nichols et al. (2007) definition for Psi2.

[darryl]

As a side note, if you are using the MSOccupancy model be aware that Psi1 and Psi2 are defined differently by Nichols et al (2007) (also as in MARK) than what is described on the occupancy web site. This may cause confusion if you are trying to duplicate results with RMark/MARK and the exercise on the website. In Nichols et al (2007) Psi2 is a conditional probability such that the probability of being in state 2 is Psi1*Psi2 whereas in the occupancy web site example, Psi2 is the probability of being in state 2 which leads to the necessary constraint that Psi1+Psi2<=1. This constraint is not needed with the Nichols et al. (2007) definition for Psi2.

For anyone wanting more detail on the multi-state occupancy (MSOccupancy) parameterizations I've just posted a note on this in the PRESENCE section of phidot. I'd also point out that is the parameterization given in our book and the online exercises was implemented in MARK or PRESENCE, then the mlogit link function could be used to ensure that psi1+psi2 <=1.

[cooch]

I've loaded RMark 1.7.4 on the website. Of particular note here is that the source (RMarkSource.zip) can now be downloaded and used on a Linux machine. The source can be found at the bottom of the RMark download page. At this point if you want to use RMark on a Linux machine you'll have to get mark.exe from Evan.

As per Jeff's note, Gary and I have figured out how to compile the 'numerical part' of MARK (mark.exe on a Windows machine) for Linux (testimony to the quality of Gary's code is that it compiled with only a few required tweaks - egc, pers. obs).

As such, for those of you who (i) have a Linux box, (ii) don't want to run Windows MARK under Wine or VMWare (although that works fine), and (iii) are comfortable (or getting there) with RMark (and can get it compiled on your Linux box), you're welcome to grab a version of the compiled MARK numerical routine(s). At present, I've made a 32-bit and 64-bit compile of the code, as 'platform-neutral' as possible. If you're running a 32-bit distro on a 32-bit box, then grab the 32-bit version. 64-bit version, otherwise.

After changing permissions to make the file executable (if you don't know how, you should probably stop now...), simply execute the file

./mark32

or

./mark64

If it returns something looking like the following, you're in business:

Code: Select all

 Program  MARK  - Survival Rate Estimation with Capture-Recapture Data
   gfortran(Linux) Vers. 5.0 Dec 2007   14-Jan-2008 13:46:32    Page  001

 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
 
     Time in minutes for this job was 0.00
     Time in minutes for this job was 0.00

          E X E C U T I O N   S U C C E S S F U L


Note:

1) to get either mark32, or mark64, send me a note offline.

2) if things don't work - don't bother Gary with it. The Linux stuff sits in my hands (for better or worse). If you can't get it to run, then send me full details (distro, type of CPU in machine, etc). The easiest thing to to is run

Code: Select all

cat /proc/cpuinfo
cat /proc/version


and email me the output...

3) In future, the Linux version will always be slightly behind the Windows version in terms of the models it handles. At present, the Linux version is identical to the Windows version bundled with MARK 5.0.

4) in rendering these compilations 'platform-neutral', I've sacrificed 'optimizations' for 'portability' - meaning, it might not run as fast as it would if I turned on all the compiler optimizations that work. Since doing so is entirely platform dependent, and since I only have a few platform options here at my disposal to test on, I may or may not be able to do an optimized compile for you.