Skip to first unread message

Scarlett Taylor

Sep 16, 2021, 12:49:58 PM9/16/21
to distance-sampling
Hi there, 

I'm wondering if there is a way to get WAIC scores instead of AIC scores from Distance in R for model selection with MCDS? 


Scarlett Taylor  (she/her)
MSc. Candidate
Integrated Fisheries Lab | Department of Biology | LSC 7088
Dalhousie University | Halifax, NS B3H 4R2

Halifax Coordinator: 
Sharks of the Atlantic Research and Conservation Centre (ShARCC)

e: scarlet...@dal.ca
c: (905) 872-8534

Eric Rexstad

Sep 16, 2021, 2:12:19 PM9/16/21
to Scarlett Taylor, distance-sampling

WAIC is not computed by the ds()​function.  There is a function AIC()​ in the Distance package that can compute Bayes Information Criterion (BIC) by setting the argument k​ to be the natural log of the number of detections.

Example here if modelobj​ is the object returned by ds()​ and nval​ is the number of detections.
BIC <- AIC(modelobj, k = log(nval))$AIC
As I understand WAIC, it requires a posterior distribution of parameters, hence is Bayesian in nature.  The Distance package does not carry out Bayesian estimation.

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Scarlett Taylor <scarlet...@hotmail.com>
Sent: 16 September 2021 17:49
To: distance-sampling <distance...@googlegroups.com>
Subject: {Suspected Spam} [distance-sampling] WAIC?
You received this message because you are subscribed to the Google Groups "distance-sampling" group.
To unsubscribe from this group and stop receiving emails from it, send an email to distance-sampl...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/c3053b89-c505-4474-8f23-d209968ec10dn%40googlegroups.com.

Scarlett Taylor

Sep 16, 2021, 2:27:23 PM9/16/21
to distance-sampling
Hi Eric, 

Thanks for confirming my suspicions. That's what I was thinking, but I wanted to confirm it before going back to my supervisor and telling them it wasn't possible. It's nice to know that you can calculate BIC however! 

Reply all
Reply to author
0 new messages