interpreting abundance estimates

[Patrick Taggart]

Hello again,

I am hoping I may be able to find some help with interpreting my abundance estimates from the model below. I am modelling the abundance of cats using occuRN and have included location as a covariate on abundance and camera as a covariate on detection. From what I understand the abundance estimates I have obtained below represent the relative abundance of cats per site at each of my two locations. I have been told not to interpret these values (14.78 on KI and 1.30 on mainland) as absolute abundance per site and that they more correctly represent the magnitude of difference in abundance between KI and mainland. Is this correct? and if so does that mean that I can divide one by the other to get a ratio of cats per site on KI : cats per sites on mainland? This ratio would be more useful than the current estimates if they cannot be interpreted as absolute abundance per site.

Thanks in advance,

Pat

Model 6

Call:

occuRN(formula = ~camera ~ ki.mainland, data = final)

Abundance (log-scale):

             Estimate    SE     z  P(>|z|)

(Intercept)      2.69 0.342  7.88 3.30e-15

ki.mainland2    -2.43 0.512 -4.74 2.12e-06

Detection (logit-scale):

            Estimate    SE      z  P(>|z|)

(Intercept)   -4.018 0.629 -6.388 1.68e-10

camera2        1.031 0.622  1.658 9.73e-02

camera3        1.734 0.614  2.823 4.76e-03

camera4        0.594 0.644  0.923 3.56e-01

AIC: 158.8067 

Number of sites: 22

optim convergence code: 0

optim iterations: 48 

Bootstrap iterations: 0 

 

 

Abundance estimate

Predicted      SE             lower          upper          ki.mainland

14.781568      5.0528655      7.5639660      28.886270      ki

1.304319       0.7477629      0.4240256      4.012135      mainland

[John Clare]

Pat,

Your interpretation of the estimate depends upon how well you think/can justify that the explicit/implicit assumptions of the model are correct or how robust the model is to assumption violations. If you interpret N as relative abundance because you think the model is not completely reasonable, then for the ratio of the two estimates to be correct if the estimates are wrong requires that the model is biased in a very particular scalar way. 

More practically, this paper (http://onlinelibrary.wiley.com/doi/10.1111/1365-2664.12883/full) compares RN estimates to SCR estimates of density (which may or may not be correct either, but relax certain assumptions). 

John