Detectability parameters for polygon detectors

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

Detectability parameters for polygon detectors

Postby lenasmarques » Mon Feb 27, 2017 4:00 pm

Hello all:

I have two independent datasets of voles’ detection data, one from live-trapping and the other from non-invasive genetics of fecal samples. The main objective of this study was to compare both methods to evaluate the use of noninvasive genetics to estimate population densities of small mammals.

Both datasets were analyzed with secr independently, the trap data using a model for multi-catch traps and the noninvasive genetics data using polygon detectors with detections modeled as counts, both with exponential detection functions.

Other than comparing the density estimates obtained from each method, i would also like to compare detectability parameters, especially g0, but i am not confident if this comparison can be made in such a direct manner (as in detectability is more than 10 times higher for noninvasive genetics (g0 = 0.07) than for trapping data (g0 = 0.0036)). If this straightforward comparison cannot be made, is there any way to make the detection parameters comparable between both methods?

Thanks in advance,
Helena
lenasmarques
 
Posts: 2
Joined: Wed Feb 22, 2017 6:54 pm

Re: Detectability parameters for polygon detectors

Postby murray.efford » Tue Feb 28, 2017 5:52 am

Hello Helena
For myself, I would be very uneasy about making a direct comparison of g0 - what would it mean? At least comparison of sigma could be construed as a comparison of trap-revealed and faecal-revealed home ranges (subject to qualification because neither behaviour - defaecation or tendency to enter a trap - is necessarily uniform across the home range). Comparisons of g0 (or better lambda0, using detectfn = 'HHN') to my mind only make sense when combined with sigma. You could devise an overall measure of detection efficiency (expected no. captures per unit effort, a function of lambda0 and sigma together) and compare that between methods. I can see that standardising 'effort' would be hard for two such radically different methods (and that is part of the problem in direct comparison of g0).
I hope this is some help - I know it stops short of giving the formula for expected number of captures - I may do that later if you want to go down that path.
Murray
murray.efford
 
Posts: 482
Joined: Mon Sep 29, 2008 7:11 pm
Location: Dunedin, New Zealand

Re: Detectability parameters for polygon detectors

Postby lenasmarques » Tue Mar 21, 2017 12:59 am

Hello Murray:

Thank you for your speedy response. It was indeed helpful, at least to be aware of what I can’t do. I’ve been doing some re-reading of the bibliography, and it really doesn’t make much sense to directly compare g0 between both methods.

The initial idea was really to be able to compare detection efficiency. Our final models give us higher vole densities with the noninvasive genetics dataset. We think that’s because noninvasive genetics is more efficient in detecting voles, and it would be nice to have some metrics to confirm this. Anyway, this is not essential for us at the moment. We can just discuss our results in a different perspective.

But I have another question. You mentioned detection function HHN. I’ve used the exponential detection function in the models for the polygon detectors, because AICc of the basic model was lower when using the exponential in comparison with using the hazard exponential. Do you still advise to use the hazard exponential in this case? Or is it better to stick with the exponential? And if I continue to use the exponential, g0 still gives me a detection probability, or the expected number of captures?

Thanks,
Helena

(sorry for the long delay of my response)
lenasmarques
 
Posts: 2
Joined: Wed Feb 22, 2017 6:54 pm


Return to analysis help

Who is online

Users browsing this forum: No registered users and 1 guest

cron