JiscMail Logo
Email discussion lists for the UK Education and Research communities

Help for DISTANCE-SAMPLING Archives


DISTANCE-SAMPLING Archives

DISTANCE-SAMPLING Archives


DISTANCE-SAMPLING@JISCMAIL.AC.UK


View:

Message:

[

First

|

Previous

|

Next

|

Last

]

By Topic:

[

First

|

Previous

|

Next

|

Last

]

By Author:

[

First

|

Previous

|

Next

|

Last

]

Font:

Proportional Font

LISTSERV Archives

LISTSERV Archives

DISTANCE-SAMPLING Home

DISTANCE-SAMPLING Home

DISTANCE-SAMPLING  December 2009

DISTANCE-SAMPLING December 2009

Options

Subscribe or Unsubscribe

Subscribe or Unsubscribe

Log In

Log In

Get Password

Get Password

Subject:

Re: Model-averaging across detection functions with similar AIC/model fit characteristics

From:

Jeff Laake <[log in to unmask]>

Reply-To:

Jeff Laake <[log in to unmask]>

Date:

Fri, 4 Dec 2009 14:33:32 -0800

Content-Type:

text/plain

Parts/Attachments:

Parts/Attachments

text/plain (139 lines)

> I'm not sure I fully understand your alternative suggestion. I need to 
> go get B&A 2nd edition from the library (have only 1st edition on my 
> shelf). Won't this require having AICc model weights available? I 
> haven't seen anywhere in DISTANCE output that these are provided.
>
Not sure if the model weights are output in Distance but if not they are 
also easily calculated from the delta aic values. For each model you 
calculate exp(-.5*delta AICc) (this is exp(0)=1 for your best model) and 
then you normalize by dividing each by the sum across all of the models 
you fitted and used to create the delta AICc values.

Model averaging is quite easy to do.

--jeff



> Best regards,
>
>
> Rich Harris
> Department of Ecosystem and Conservation Sciences
> University of Montana
> Contact:
> 2175 11th St W, Missoula, MT USA 59801
> Tel/Fax: 1 (406) 542-6399
> Mobile: (406) 214-7484
> [log in to unmask]
> http://www.cfc.umt.edu/PersonnelDetail.aspx?id=1152
> http://www.bearbiology.com/ursus-journal/about-ursus-journal.html
>
>
> ----- Original Message ----- From: "Jeff Laake" <[log in to unmask]>
> To: "Rich Harris" <[log in to unmask]>
> Cc: <[log in to unmask]>
> Sent: Thursday, December 03, 2009 5:44 PM
> Subject: Re: [DIST-SAMP] Model-averaging across detection functions 
> with similar AIC/model fit characteristics
>
>
>> Rich-
>>
>> Distance does not have any model averaging capability at present but 
>> you can include model uncertainty in the variance estimate by running 
>> each model in each bootstrap and having it select the best model for 
>> each bootstrap run.  To do this you put each model specification in 
>> the analysis specification and tell it to select based on min aic.  
>> Now the only oddity in all that is you are still left with using the 
>> point estimate from the best model unless you want to take the 
>> average density across all the bootstrap runs.
>>
>> But given the problems you are having with the bootstrap with the 
>> small number of observations, I suggest you take the estimates and 
>> variances from your various models and construct the model averaged 
>> estimate and its variance.  It is quite easy to do and probably 
>> should be included in Distance (what say you Len?).  Note that when 
>> you compute the variance for the model averaged estimate, don't use 
>> the formula 4.9 in B&A model selection.  Instead use 6.12 on page 345 
>> of the second edition.
>>
>> --jeff
>>
>> Rich Harris wrote:
>>> Distance Samplers:
>>>
>>> This is probably perfectly obvious, but this old dog is having 
>>> trouble learning the new tricks. I thought that specifying multiple 
>>> detection functions within the CDS analysis engine via the 'Models' 
>>> tab (with, e.g., AIC as the selector) would yield a model average 
>>> point estimate as well as variances as estimated using the 
>>> bootstrapping. I'm getting no end of error messages from the 
>>> bootstrapping, and although I do, in the end, get bootstrapped 
>>> variance estimates for some (not all) of the selected models, the 
>>> remainder of the output is identical to that of the top model (as 
>>> determined in a separate, stand-alone) run. In one-at-a-time runs, 
>>> reasonable detection functions all have similar AIC, log-likelihood, 
>>> and GOF properties, but return rather different estimates of 
>>> density, thus my discomfort with using only the top model. Sample 
>>> size is, alas, small (32 total animal groups, 30 after 5% 
>>> truncation, 29 after 10%), and this may be the root of the problem. 
>>> Alas, the animals are in Mongolia and this survey is not easily 
>>> replicated - the sample size is what it is! (Also aggravating is 
>>> that all binning and truncation strategies tend to lead, naively, 
>>> toward the negative exponential, which I'm uncomfortable relying on, 
>>> so am tentatively rejecting it despite it AIC-appeal!).
>>>
>>> Any clues to where I'm going wrong?
>>>
>>> Thanks!
>>>
>>> Rich Harris
>>> Department of Ecosystem and Conservation Sciences
>>> University of Montana
>>> Contact:
>>> 2175 11th St W, Missoula, MT USA 59801
>>> Tel/Fax: 1 (406) 542-6399
>>> Mobile: (406) 214-7484
>>> [log in to unmask]
>>> http://www.cfc.umt.edu/PersonnelDetail.aspx?id=1152
>>> http://www.bearbiology.com/ursus-journal/about-ursus-journal.html
>>>
>>> -- 
>>> distance-sampling list home page:
>>>  http://www.jiscmail.ac.uk/lists/distance-sampling.html
>>> To unsubscribe, go to the home page and click "Join or leave the list",
>>>  or email [log in to unmask] with the following message body:
>>>  leave distance-sampling
>>> Before posting to the list, read the list guidelines at
>>>  http://www.jiscmail.ac.uk/files/distance-sampling/
>>>    distance-sampling.welcome
>>> Any questions, email the list owner at
>>>  [log in to unmask]
>>> -- 
>
> -- 
> distance-sampling list home page:
>  http://www.jiscmail.ac.uk/lists/distance-sampling.html
> To unsubscribe, go to the home page and click "Join or leave the list",
>  or email [log in to unmask] with the following message body:
>  leave distance-sampling
> Before posting to the list, read the list guidelines at
>  http://www.jiscmail.ac.uk/files/distance-sampling/
>    distance-sampling.welcome
> Any questions, email the list owner at
>  [log in to unmask]
> -- 

--
distance-sampling list home page:
  http://www.jiscmail.ac.uk/lists/distance-sampling.html
To unsubscribe, go to the home page and click "Join or leave the list",
  or email [log in to unmask] with the following message body:
  leave distance-sampling
Before posting to the list, read the list guidelines at
  http://www.jiscmail.ac.uk/files/distance-sampling/
    distance-sampling.welcome
Any questions, email the list owner at
  [log in to unmask]
--

Top of Message | Previous Page | Permalink

JiscMail Tools


RSS Feeds and Sharing


Advanced Options


Archives

July 2014
June 2014
May 2014
April 2014
March 2014
February 2014
January 2014
December 2013
November 2013
October 2013
September 2013
August 2013
July 2013
June 2013
May 2013
April 2013
March 2013
February 2013
January 2013
December 2012
November 2012
October 2012
September 2012
August 2012
July 2012
June 2012
May 2012
April 2012
March 2012
February 2012
January 2012
December 2011
November 2011
October 2011
September 2011
August 2011
July 2011
June 2011
May 2011
April 2011
March 2011
February 2011
January 2011
December 2010
November 2010
October 2010
September 2010
August 2010
July 2010
June 2010
May 2010
April 2010
March 2010
February 2010
January 2010
December 2009
November 2009
October 2009
September 2009
August 2009
July 2009
June 2009
May 2009
April 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
January 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
February 2007
January 2007
2006
2005
2004
2003
2002
2001
2000
1999
1998


JiscMail is a Jisc service.

View our service policies at https://www.jiscmail.ac.uk/policyandsecurity/ and Jisc's privacy policy at https://www.jisc.ac.uk/website/privacy-notice

Secured by F-Secure Anti-Virus CataList Email List Search Powered by the LISTSERV Email List Manager