Different density estimates with data subsets in secr

questions concerning anlysis/theory using program DENSITY and R package secr. Focus on spatially-explicit analysis.

Different density estimates with data subsets in secr

Postby dlm122 » Sun Jun 02, 2013 11:49 am

Hello,

I am analyzing DNA data, which are collections of chimpanzee fecal material from searches across 624 sq km. I am using the 'polygon' detector type, and my polygons are 4x4 km grid squares that I drew over my searched area. I obtained a population density estimate of 0.25 (0.16-0.37) individuals/sq. km (perfectly reasonable for this savanna-woodland habitat). My complete data set includes 113 individuals, 38 recaptures, and 49 detectors used, 17 visited.

Previously, researchers had designated part of the region where I worked "suitable" chimpanzee habitat (presumably, the remaining areas "not so suitable"). In the interest of further investigating these designations, I divided my captures and traps data files into "suitable" and "unsuitable", and calculated the density estimates for each of these subsets. I obtained population density estimates of 0.39 (0.16-0.37) and 0.32 (0.22-0.46) for "suitable" and "unsuitable", respectively.

Clearly, the second analyses had smaller sample sizes, but would this be the sole factor driving the different results? For the second analyses, I had the following variables:

SUITABLE: 58 individuals, 12 recaptures, 30 detectors (12 visited)
UNSUITABLE: 55 individuals, 26 recaptures, 18 detectors (6 visited)

Also, does the number of detectors (i.e., my grid squares), which is randomly decided in this case, matter?

Thank you very much for your help.
dlm122
 
Posts: 8
Joined: Wed Oct 31, 2012 12:08 pm

Re: Different density estimates with data subsets in secr

Postby murray.efford » Sun Jun 02, 2013 7:03 pm

It's difficult to be definite on the information you have given, and you must have mis-transcribed one of the estimates. The intervals overlap widely, and results will jump around as samples get smaller, but it is possible something may have gone awry in the analysis - perhaps an inadequate integration buffer? The number of detector polygons (grid squares) is not a factor in itself. The subdivided analysis (suitable vs not so suitable habitat) would be better done using a categorical mask covariate - more elegant, and more logical as it's likely the 'catchment area' of a particular searched polygon can include both types of habitat (Efford & Fewster 2013 show a skink example).
Murray
murray.efford
 
Posts: 686
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand

Re: Different density estimates with data subsets in secr

Postby dlm122 » Mon Jun 03, 2013 11:38 am

Thanks so much for the reply. I did try several different buffer widths, with almost identical results. But I will also try a categorical mask covariate.

p.s. sorry - suitable results were 0.39 (0.20 - 0.73)
dlm122
 
Posts: 8
Joined: Wed Oct 31, 2012 12:08 pm


Return to analysis help

Who is online

Users browsing this forum: No registered users and 18 guests