numeric underflow

questions concerning analysis/theory using program MARK

numeric underflow

Postby sixtystrat » Mon Aug 06, 2018 2:27 pm

I am running a Pradel model with random effects on p for a robust design data set spanning 12 years with 3-8 annual capture occasions. My estimates for phi, sigmap, p, and c are all reasonable with good standard errors but I get a "numerical underflow" warning. This happens even when the model is heavily constrained (dot models). The message goes away when I set sigmap=0. What causes this and should I be concerned? Thanks!!!
Joe
sixtystrat
 
Posts: 115
Joined: Wed Jun 01, 2011 9:19 am

Re: numeric underflow

Postby cooch » Mon Aug 06, 2018 3:16 pm

No - that error isn't a huge issue (shows up now with some frequency when MARK's codebase was 'updated' to higher precision numerics. The 'failed to coverge warning' is a real issue, but 99/100 times, the 'underflow' warning can be safely ignored.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: numeric underflow

Postby sixtystrat » Tue Aug 07, 2018 9:40 am

Perfect! Thanks Evan.
Joe
sixtystrat
 
Posts: 115
Joined: Wed Jun 01, 2011 9:19 am

Re: numeric underflow

Postby sixtystrat » Tue Aug 07, 2018 12:40 pm

One other thing, where do I find the Fletcher's c-hat? It does not seem to be listed in the output in notepad. Thanks.
sixtystrat
 
Posts: 115
Joined: Wed Jun 01, 2011 9:19 am

Re: numeric underflow

Postby cooch » Fri Nov 02, 2018 3:21 pm

sixtystrat wrote:One other thing, where do I find the Fletcher's c-hat? It does not seem to be listed in the output in notepad. Thanks.


Its in the 'full output' -- button just to the right of the 'trash can' in the MARK toolbar.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University

Re: numeric underflow

Postby sixtystrat » Mon Nov 05, 2018 4:07 pm

Thanks Evan. So has the chat listed in the output already been adjusted ala Fletcher(2012)?
sixtystrat
 
Posts: 115
Joined: Wed Jun 01, 2011 9:19 am

Re: numeric underflow

Postby cooch » Mon Nov 05, 2018 4:11 pm

sixtystrat wrote:Thanks Evan. So has the chat listed in the output already been adjusted ala Fletcher(2012)?


Yes, one and the same.
cooch
 
Posts: 1628
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University


Return to analysis help

Who is online

Users browsing this forum: No registered users and 9 guests