Different output with newest version

announcements (new versions, changes, bugs, installation problems...) related to program PRESENCE

Different output with newest version

Postby vxrojas » Sat Aug 28, 2021 7:22 pm

I recently downloaded the newest version of Presence (13.12), and opened up a .pa3 file from a previous project I ran in 2018 on version 11.3. Everything opened fine, my data looked ok, and the last models I ran were showing up just as I had left them. When I go to run a new model, I noticed I was getting SEs of 0 for everything. So I ran a previous model, and just changed the name. I used the retrieve model option to make sure it was exactly the same. With the output, I get the same values for everything (AIC, w, parameter estimates, etc) except the SEs and CIs (all set so that SEs=0).

I looked more closely and noticed a couple of different things with the output. In v11.3, I get this:
********* Input Data summary *******
Number of sites = 162
Number of sampling occasions = 4
Number of states = 0
Number of missing observations = 0
Data checksum = 15524


But in v13.12, I get this:
********* Input Data summary *******
Number of sites = 162
Number of sampling occasions = 4
Number of states = 2
Number of missing observations = 0



The number of states have changed, both were ran as simple single season.

Other different output in case that's helpful:

v11.3
Untransformed Estimates of coefficients for covariates (Beta's)
======================================================================
estimate std.error
A1 psi : -0.354876 0.750110
A2 psi.for : 0.957810 0.422500
B1 P[1] : 7.145611 2.476664
B2 P[1].date : -1.091891 0.290107


v13.12
Untransformed Estimates of coefficients for covariates (Beta's)
======================================================================
estimate std.error
A1 psi.a1 : -0.354876 0.000000
A2 psi.for : 0.957810 0.000000
B1 P[4].b1 : 7.145611 0.000000
B2 P[1].date : -1.091891 0.000000


and also this statement that I have not seen in the previous version:
attempt 0/0 (iv= 0.0000 0.0000 0.0000...), -2loglike= 200.4056, optmiz time:0.0 min.
0 random initial value vectors attempted:
Initial values:
0.0000 0.0000 0.0000 0.0000 0.0000


Apologies if this is not a software/glitch issue, but I was wondering if something has changed between the versions where perhaps now I have to change how I enter/run the models, or if there is just some sort of error happening. I tried running other pa3 files from the previous version and I'm getting the same issue. Thanks in advance for any guidance.
vxrojas
 
Posts: 2
Joined: Sat Aug 28, 2021 5:11 pm

Re: Different output with newest version

Postby jhines » Mon Aug 30, 2021 8:56 am

I have inadvertently clicked the checkbox labelled, "Don't compute V-C matrix" at times and the result is that the standard errors are printed as zeros. Is it possible you've done the same? If not, please send me your most recent backup zipfile or pao file and I will try to find out what is going on.

Jim
jhines
 
Posts: 527
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Different output with newest version

Postby jhines » Mon Aug 30, 2021 9:03 am

The other output you mentioned is a new feature where Presence will try multiple sets of random initial values to the optimization function in order to prevent converging on local maxima. The first set of initial values is a vector of all zeros, as you saw in your output. If you checked the box, "Run with random initial values", Presence will ask how many random initial value vectors to try, then try the optimization with that many initial value vectors. This is not usually needed for simple single-season models, but I've seen a few cases where it was needed, so it might be a good idea to try a few dozen.
jhines
 
Posts: 527
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Different output with newest version

Postby jhines » Mon Aug 30, 2021 11:09 am

Thanks for sending the file. I found the problem... one of those things I didn't anticipate.

The checkbox options are passed to the Presence exe file via character strings. Although you didn't specify the option to not compute standard errors, the string passed to the exe file for that option is "nose", which appears in the pathname to the exe file (c:\users\asdf\downloads\presence_nosetup\presence.exe). The solution is to rename the folder from:

c:\users\asdf\downloads\presence_nosetup

to

c:\users\asdf\downloads\presence

I will modify the next version (2.13.13) to handle this situation.
jhines
 
Posts: 527
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Different output with newest version

Postby vxrojas » Sat Sep 11, 2021 7:50 pm

Thanks so much, everything is working and SEs are no longer 0. I appreciate how quickly you responded and resolved my dilemma.
vxrojas
 
Posts: 2
Joined: Sat Aug 28, 2021 5:11 pm


Return to software problems/news

Who is online

Users browsing this forum: No registered users and 0 guests