model averaging - parameter matrix not correct length

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

model averaging - parameter matrix not correct length

Postby Lyndon » Wed Jun 15, 2016 11:42 pm

Hi,
I've obtained model averaged estimates numerous time without problems. Today I wanted to change the c-hat and obtain new estimates. Now I'm receiving an error message saying 'Parameter Matrix not the correct length in output: f0 1: Session x'. This happened for both a CRD model and a MSCRD model. All parameters are OK except for abundances. While specifying, where I expect a tab for 'N Site Session', I'm getting 'Never encountered (f0) 1: Site Session'. The individual model estimates are fine. I downloaded and installed the latest version of Mark but got the same result.
Has anyone encountered this problem or know a fix?
Thanks,
Lyndon.
Lyndon
 
Posts: 16
Joined: Thu May 12, 2011 9:57 pm

Re: model averaging - parameter matrix not correct length

Postby egc » Thu Jun 16, 2016 7:40 am

Lyndon wrote:Hi,
I've obtained model averaged estimates numerous time without problems. Today I wanted to change the c-hat and obtain new estimates. Now I'm receiving an error message saying 'Parameter Matrix not the correct length in output: f0 1: Session x'. This happened for both a CRD model and a MSCRD model. All parameters are OK except for abundances. While specifying, where I expect a tab for 'N Site Session', I'm getting 'Never encountered (f0) 1: Site Session'. The individual model estimates are fine. I downloaded and installed the latest version of Mark but got the same result.
Has anyone encountered this problem or know a fix?
Thanks,
Lyndon.


This may be one of the parts of MARK that didn't get updated when parameterization was changed from N -> f0. Gary will need to confirm/fix.
egc
Site Admin
 
Posts: 201
Joined: Thu May 15, 2003 3:25 pm

Re: model averaging - parameter matrix not correct length

Postby gwhite » Thu Jun 16, 2016 8:17 am

This does indeed sound like an issue caused by changing the parameterization from N to f0. I would need the MARK files to figure out exactly what is happening. I would suggest you re-run the models and get the proper parameterization on all of them.

Gary
gwhite
 
Posts: 329
Joined: Fri May 16, 2003 9:05 am

Re: model averaging - parameter matrix not correct length

Postby Lyndon » Thu Jun 16, 2016 7:25 pm

Thank you Gary. I'll have a look at how one changes the parameterization. I actually didn't change anything, only adjusted c-hat. I'll have a look to see what happens when I rerun the models. If that doesn't fix it, then I'll run them again and have a good look around for changing parameterization.
Lyndon.
Lyndon
 
Posts: 16
Joined: Thu May 12, 2011 9:57 pm

Re: model averaging - parameter matrix not correct length

Postby Lyndon » Thu Jun 16, 2016 9:27 pm

OK, I've missed the point. I retrieved, renamed and reran the models. This time I got estimates but on some scale I didn't understand. How do I get the proper paramaterization? Do you want me to send you the Mark files? Lyndon.
Lyndon
 
Posts: 16
Joined: Thu May 12, 2011 9:57 pm

Re: model averaging - parameter matrix not correct length

Postby gwhite » Fri Jun 17, 2016 8:06 am

Have you looked at the derived estimates, where N estimates are reported?

You can send me the files if you are still having trouble and I'll figure out what is going on.

Gary
gwhite
 
Posts: 329
Joined: Fri May 16, 2003 9:05 am

Re: model averaging - parameter matrix not correct length

Postby Lyndon » Sun Sep 25, 2016 9:47 pm

Hi Gary,
Sorry I dropped the thread earlier. I ran out of time and worked out some estimates by hand.
I'm now working on a ORDMS. A discussion at an advanced workshop indicated that the last rather than the first pent was obtained by subtraction in this model. Can you confirm that?
I'm using version 8.1.
BTW, the model-averaging problem I had with my MSORD isn't occurring for the ORDMS.
Thanks,
Lyndon.
Lyndon
 
Posts: 16
Joined: Thu May 12, 2011 9:57 pm

Re: model averaging - parameter matrix not correct length

Postby Lyndon » Sun Sep 25, 2016 9:53 pm

Model averaging works fine for derived estimates as you suggested BTW.
Lyndon
 
Posts: 16
Joined: Thu May 12, 2011 9:57 pm

Re: model averaging - parameter matrix not correct length

Postby cooch » Sun Sep 25, 2016 10:11 pm

Lyndon wrote:Hi Gary,
Sorry I dropped the thread earlier. I ran out of time and worked out some estimates by hand.
I'm now working on a ORDMS. A discussion at an advanced workshop indicated that the last rather than the first pent was obtained by subtraction in this model. Can you confirm that?


Right out of the MARK helpfile:

pent(s, t) - probability of entry onto the study area of state s at time t. The probability of entering just prior to the last secondary occasion is obtained by subtraction, so that the sum of the pent values in a primary session should be constrained to <=1 using the MLogit link function. An example of how to invoke constraints and still use the MLogit link function is also provided.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: model averaging - parameter matrix not correct length

Postby Lyndon » Wed Sep 28, 2016 11:22 pm

OK, thanks. Employed the MLogit constraint OK, it's just that the estimate obtained by subtraction makes much more sense if it were the first, prior to the first sample (as in POPAN), rather than the last, after the last sample.
Lyndon
 
Posts: 16
Joined: Thu May 12, 2011 9:57 pm

Next

Return to software problems/news

Who is online

Users browsing this forum: No registered users and 12 guests