SD/parameter issues?

posts related to the RMark library, which may not be of general interest to users of 'classic' MARK

SD/parameter issues?

Postby j.harv3y » Fri Oct 19, 2018 10:17 am

Hi,

Apologies for the vagueness, but I just can't work out what the problem is. I've read the book, and tried the suggested fixes in c.10 but nothing seems to be working!

When I run the model with the lowest AICC I get the following output;

Output summary for CJS model
Name : Phi(~time)p(~1)

Npar : 6 (unadjusted=4)
-2lnL: 249.2936
AICc : 261.8186 (unadjusted=257.54051)

Beta
estimate se lcl ucl
Phi:(Intercept) 20.2865380 0.0000000 20.2865380 20.2865380
Phi:time2013 -15.9651950 0.0000000 -15.9651950 -15.9651950
Phi:time2014 -0.3592642 0.0000000 -0.3592642 -0.3592642
Phi:time2015 -19.1239460 0.0000000 -19.1239460 -19.1239460
Phi:time2016 -18.4737820 0.0000000 -18.4737820 -18.4737820
p:(Intercept) 1.2695096 0.2162058 0.8457463 1.6932729


Real Parameter Phi

2012 2013 2014 2015 2016
2012 1 0.9868921 1 0.7618035 0.8596947
2013 0.9868921 1 0.7618035 0.8596947
2014 1 0.7618035 0.8596947
2015 0.7618035 0.8596947
2016 0.8596947


Real Parameter p

2013 2014 2015 2016 2017
2012 0.7806588 0.7806588 0.7806588 0.7806588 0.7806588
2013 0.7806588 0.7806588 0.7806588 0.7806588
2014 0.7806588 0.7806588 0.7806588
2015 0.7806588 0.7806588
2016 0.7806588


If I set adjustment=FALSE, I get the exact same output however with 4 parameters being run.

After looking at Model.3$results$singular I get the following;
[1] 5 3

and if I then change the singular b's, and refit the model, the SD is still the same (0)- why?!!

I looked through the capture histories and there are no individuals who were captured only on 2012 and 2014 and also 2012 and 2016 (confirming the both 3&5 are at the boundary).

I have also tried using list(formula=~-1 + time), but with no luck.

Is this the issue, and if so how should I approach 'fixing' it?

I'd really appreciate any advice/suggestions.

Thanks so much in advance
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am

Re: SD/parameter issues?

Postby jlaake » Fri Oct 19, 2018 12:09 pm

No one said there is anything wrong. You just have 2 parameters at boundary. Not sure if it was in the material you read but when parameters are at boundaries the std errors can be unreliable. One option is to get profile confidence intervals. Another is to try formula=~-1+time, link="sin" for parameter specification. The sin link handles boundary parameters better than the default logit link but it can only be used with an identity design matrix which you get when you use -1+time.
jlaake
 
Posts: 1417
Joined: Fri May 12, 2006 12:50 pm
Location: Escondido, CA

Re: SD/parameter issues?

Postby j.harv3y » Thu Oct 25, 2018 12:36 pm

Thanks so much!!
Problem solved, you're amaaazing!
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am


Return to RMark

Who is online

Users browsing this forum: No registered users and 7 guests