modeling subsets of all variables

76 views
Skip to first unread message

Jason Unruh

unread,
Jan 7, 2015, 1:50:51 PM1/7/15
to unma...@googlegroups.com
I am using the pcount function and I have 5 variables (some continuous, some categorical) that may influence detection probability for 9 species of birds.  I want to model which variables give the best model for detection probability.  Is there a way to easily model and compare all subsets/combinations of my variables, or do I have to write every single combination out separately? 

Detection Variables: 
observer
date
time
visit
wind


Kery Marc

unread,
Jan 7, 2015, 5:36:12 PM1/7/15
to unma...@googlegroups.com
Dear Jason,

I believe that you can use functionality in Marc Mazerolle's AICcmodavg package to do what you want.

Kind regards  ----  Marc


From: unma...@googlegroups.com [unma...@googlegroups.com] on behalf of Jason Unruh [jason...@gmail.com]
Sent: 07 January 2015 19:50
To: unma...@googlegroups.com
Subject: [unmarked] modeling subsets of all variables

--
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Marc Mazerolle

unread,
Jan 8, 2015, 9:07:31 AM1/8/15
to unma...@googlegroups.com
Hi Jason,

If you want to compare different models, you can use some of the functionalities of the AICcmodavg package (model selection, multimodel inference, check GOF), as Marc Kéry suggested. However, you will have to write out each model that you want to compare. I intentionally did not include any functions to automagically generate all possible combinations of variables to steer users away from "fishing expeditions" and keep the number of models down. It is important to think about each model you want to include in your candidate model set and check the output of each model for potential problems (e.g., lack of convergence, parameters on boundaries, ill conditioning). Just looking at the resulting model selection table to see which model is best can be misleading.

Hope it helps,

Marc

------------------------------
Centre d'étude de la forêt
Université du Québec en Abitibi-Témiscamingue
445 boulevard de l'Université
Rouyn-Noranda, Québec J9X 5E4, Canada
Tel: (819) 762-0971 ext. 2458
Email: marc.ma...@uqat.ca

Jason Unruh

unread,
Jan 8, 2015, 12:42:54 PM1/8/15
to unma...@googlegroups.com
Thanks, those are helpful suggestions.
Reply all
Reply to author
Forward
0 new messages