in pcountOpen, using obsCovs to estimate lambda, gamma, or omega?

347 views
Skip to first unread message

Rhine

unread,
Aug 2, 2012, 2:03:29 PM8/2/12
to unma...@googlegroups.com

Is it possible to use obsCovs to estimate lambda, gamma, or omega? For example, I have been playing with the alfl0506 data and pcountOpen script from the January Webinar. When I try to include time as a covariate to help estimate lambda like this:

 

(fm5 <- pcountOpen(~time, ~1, ~1, ~1, alfl.umf, K=30))

 

I get this error:

in model.frame.default(lamformula, siteCovs, na.action = NULL) :

  object is not a matrix

 

The reason I ask is that I am hoping to use monthly fruiting of tree species in three different sites as covariates to explain monthly bird counts. I don't think I can include fruiting as siteCovs because fruiting is not a constant feature of my sites – it changes each month.

 

I have attached an example spreadsheet showing how I think I would need to setup my data to use pcountOpen.  Though I've only included 6 dates, I have 18 dates in my full data set. I welcome any suggestions.

 

Thanks, Rhine

pcountopen.testdata.csv

Jeffrey Royle

unread,
Aug 2, 2012, 2:08:45 PM8/2/12
to unma...@googlegroups.com
hi Rhine,
 I hadn't realized in your previous email that you have only 3 sites. It seems unlikely to me that you will have much luck with the DM model in this case -- it really depends on having a good number of spatial replicates.
 As for the specific question related to the alfl data, I don't know whats going on there....
regards,
andy

Rhine

unread,
Aug 2, 2012, 2:39:49 PM8/2/12
to unma...@googlegroups.com
Thanks,

We actually have eight sub-plots per site which potentially could be included in our data structure. So far, we have collapsed the data from the subplots as we think that birds are moving freely among them so that the number of birds in subplot 2 could reflect fruiting in subplot 1, etc.

Would a total of 24 subplots grouped into three different sites provide enough replication for pcountOpen?

And, despite my errors with the alfl data, in principle is it possible to use obsCovs as factors when estimating lambda, gamma, or omega?

Rhine

Andy Royle

unread,
Aug 2, 2012, 2:44:38 PM8/2/12
to unma...@googlegroups.com

hi Rhine,
 24 is probably a good starting point. I can't give specific guidance on this.
 I don't believe that obsCovs should affect the main dynamics parameters but, rather, the encounter rate parameter.
regards,
andy

 
-----unma...@googlegroups.com wrote: -----

To: unma...@googlegroups.com
From: Rhine <sin...@franklinpierce.edu>
Sent by: unma...@googlegroups.com
Date: 08/02/2012 02:39PM
Subject: Re: [unmarked] in pcountOpen, using obsCovs to estimate lambda, gamma, or omega?

Rhine

unread,
Aug 2, 2012, 2:52:33 PM8/2/12
to unma...@googlegroups.com
So does this mean that if I want to look at how covariates that change with each sample period (such as fruiting) affect the main dynamics parameters, I need to designate them as siteCovs or yearlySiteCovs?

Thanks

Richard Chandler

unread,
Aug 2, 2012, 2:59:12 PM8/2/12
to unma...@googlegroups.com
Hi Rhine,

For pcountOpen(), "obsCovs" are only used for modeling detection probability when you collected data using the so-called "robust design" (which is always the preferred option). The possible covariate options for the 4 parameters are this:

lambda: siteCovs only
omega/gamma: siteCovs or yearlySiteCovs
p: siteCovs, yearlySiteCovs, or obsCovs (if you have the robust design)


Richard

_____________________________________
Richard Chandler, post-doc
USGS Patuxent Wildlife Research Center
301-497-5696



From: Rhine <sin...@franklinpierce.edu>
To: unma...@googlegroups.com
Date: 08/02/2012 02:52 PM
Subject:
Re: [unmarked] in pcountOpen, using obsCovs to estimate lambda, gamma, or omega?
Sent by: unma...@googlegroups.com


Reply all
Reply to author
Forward
0 new messages