[R-sig-ME] help with model convergence

[David Stainbrook]

To whom it may concern,

I am having some issues with model convergence with linear mixed modeling. 
I am using mixed modeling with count data (GPS locations) for 30 individuals
and landscape characteristics as covariates to estimate resource selection. 

The technique I am using is similar to those outlined in:
 
  Normal
  0
  
  
  false
  false
  false
  
   
   
   
   
   
  
  MicrosoftInternetExplorer4
 

 
 


st1\:*{behavior:url(#ieooui) }



 /* Style Definitions */
 table.MsoNormalTable
	{mso-style-name:"Table Normal";
	mso-tstyle-rowband-size:0;
	mso-tstyle-colband-size:0;
	mso-style-noshow:yes;
	mso-style-parent:"";
	mso-padding-alt:0in 5.4pt 0in 5.4pt;
	mso-para-margin:0in;
	mso-para-margin-bottom:.0001pt;
	mso-pagination:widow-orphan;
	font-size:10.0pt;
	font-family:"Times New Roman";
	mso-ansi-language:#0400;
	mso-fareast-language:#0400;
	mso-bidi-language:#0400;}





Manly, B.F.J., L.L.
McDonald, D.L. Thomas, T.L. McDonald, and W.P. Erickson. 2002. Resource
selection by animals: Statistical design
and analysis for field studies, Second Edition. Kluwer Academic Publishers,
Dordrecht.


 
  Normal
  0
  
  
  false
  false
  false
  
   
   
   
   
   
  
  MicrosoftInternetExplorer4
 

 
 


 /* Style Definitions */
 table.MsoNormalTable
	{mso-style-name:"Table Normal";
	mso-tstyle-rowband-size:0;
	mso-tstyle-colband-size:0;
	mso-style-noshow:yes;
	mso-style-parent:"";
	mso-padding-alt:0in 5.4pt 0in 5.4pt;
	mso-para-margin:0in;
	mso-para-margin-bottom:.0001pt;
	mso-pagination:widow-orphan;
	font-size:10.0pt;
	font-family:"Times New Roman";
	mso-ansi-language:#0400;
	mso-fareast-language:#0400;
	mso-bidi-language:#0400;}







Sawyer, H., Nielson, R.M.,
Lindzey, F., & McDonald, L.L. (2006) Winter habitat selection of mule deer
before and during development of a natural gas field. Journal of Wildlife
Management, 70, 396-403.



However, these resource selection methods have not incorporated mixed modeling,
which is part of what I am trying to do with my graduate thesis. Basically, I
randomly place sampling units over my 11 square mile study area (3000 circles
that are 200 meters in diameter), and then count the number of GPS locs for
each individual in those sampling units. Then, measure the landscape covariates
for each of the 3000 sampling units. Then use mixed modeling to obtain the
parameter estimates, which I then use to plug into a finer scale model across
the landscape (5 x 5 meter grid) based on each cell's covariate values.

I have attached a zipped file that includes my R script and txt file of data
for the first survey - so that you can see all of my steps. One script includes
all of the models and the other includes only model 13 which includes all of
the covariates with squared terms, and was the best model based on AIC. Each
model adds another covariate or a squared term. The more complex models take a
really long time to run (e.g. - Model 13 takes a few days to run). 

There are 7 total surveys and so far only the first survey has worked. The
models for the rest of the surveys either have a false convergence error (8) or
a function evaluation limit reached without convergence (9) error. As you can
see in my script, I have centered and normalized all of the covariates to help
with convergence and added a bit to increase the default number of iterations.
Since most count data are overdispersed, I am using the quasipoisson
distribution  - I would prefer to use negative binomial, but this is currently
not possible with lmer. I am also using an offset term to constrain my Resource
Selection Values or probabilites between 0 and 1, which are calculated as e^Y.

Are you able to help me at all with my convergence issue or offer some other
pointers? I may need to send you the datasets from the other models so you can
see why I get these errors.

Here is the current version of R and lme4 - taken from the "super" computers at
Penn State that I am running the models on - they are running linux, so it's a
bit different than on my own computer, which is running Windows XP. 

> sessionInfo()
R version 2.11.1 (2010-05-31) 
x86_64-unknown-linux-gnu 

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=C              LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] lme4_0.999375-33   Matrix_0.999375-39 lattice_0.18-8    

loaded via a namespace (and not attached):
[1] grid_2.11.1 nlme_3.1-96


Also, can you please keep this email confidential.

Thanks so much, this help will be tremendous. 

..................................................................

David P. Stainbrook
M.S. Graduate Research Assistant
Pennsylvania Cooperative Fish & Wildlife Research Unit
The Pennsylvania State University
221 Forest Resources Bldg
University Park, PA 16802
(814) 865-0772

..................................................................