Improper V-C matrix for beta estimates. Some variances non-p

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

Improper V-C matrix for beta estimates. Some variances non-p

Postby mconner » Tue Feb 06, 2018 4:09 pm

Greetings:

With Jeff's help I have managed to figure out how to bin up an age*time model for a 3 age group model (2 sex groups too, but not relevant). I can compare estimates to regurlar MARK and all seems fine. To figure out how to output model averaged values, I started easy by creating 2 models Phi(age*t) p(.) and Phi(t) p(.) as shown in the code below. The issue is when I try to model average the 2 models I get the error message:
Warning messages:
1: In get.real(model, parameter.names[i], se = se, show.fixed = show.fixed) :

Improper V-C matrix for beta estimates. Some variances non-positive.

2: In get.real(model, parameter.names[i], se = se, show.fixed = show.fixed) :

Improper V-C matrix for beta estimates. Some variances non-positive.

When I look at the VC for the betas in regular P MARK there are no non-positive varinaces. In fact, all looks hunky-dory with no weird estimates or SEs. What am I missing.

And thanks for your help already (esp you Jeff). Cheers, Mary

*****************************************************************************************
SIE_cjs.ddl=add.design.data(SIE_cjs.process,SIE_cjs.ddl,"Phi",type="age",
bins=c(0,1,2,27),
right=FALSE,name="ageclass",replace = T)

Phi.age.time=list(formula=~ageclass*time)
Phi.time=list(formula=~time)
p.dot = list(formula=~1)

Phi.age.t_p.dot=mark(SIE_cjs.process,SIE_cjs.ddl,model.parameters = list(Phi=list(formula=~ageclass*time),p=p.dot))
Phi.time_p.dot=mark(SIE_cjs.process,SIE_cjs.ddl,model.parameters = list(Phi=Phi.time,p=p.dot))

SIE_cjs_results=collect.models()
mod_avg=model.average(SIE_cjs_results)
mconner
 
Posts: 21
Joined: Wed Aug 24, 2005 7:29 pm
Location: Utah State University

Re: Improper V-C matrix for beta estimates. Some variances n

Postby mconner » Tue Feb 06, 2018 5:59 pm

With Jeff's help, I found this error was due to many se(beta)=0 values. I used drop=FALSE in the model.averaging function to solve; although I still got the error message, model averaging occured.

And then Jeff suggested I "To get a clearer picture you may want to make the intercept the oldest age class rather than using the youngest age class. You can do that with the relevel function in R"

This I will do! Thanks Jeff.
mconner
 
Posts: 21
Joined: Wed Aug 24, 2005 7:29 pm
Location: Utah State University


Return to RMark

Who is online

Users browsing this forum: No registered users and 14 guests