Interpreting 85% CIs overlap with zero for psi intercept

82 views
Skip to first unread message

haileyj...@gmail.com

unread,
Jul 29, 2023, 12:37:45 PM7/29/23
to unmarked
Hi all, 

I am running a single season single species occupancy model for an endemic chipmunk species with a sample size of 162 sites. I ran my a priori models and calculated 85% CIs for detection and occupancy. I've noticed for some of my occupancy models that the there is an overlap in zero for the psi intercept. All the occupancy variables are continuous and some also overlapped zero. How would you interpret this? 

I have two thoughts on this matter: 1) Does this mean that there is no correlation between occupancy probability and these variables? 2) Or would this problem only affect the model when all the variables are at the average value (i.e., 0 because I scaled all the variables)? 

Below is the output for a model that has psi CI overlapping zero: 

>  summary(mo140)

Call:
occu(formula = ~scale(Herbaceous_cover_near) + scale(Litter_cover_near) +
    scale(juliandate) + scale(FA_HR) + scale(Coniferall_cover) +
    scale(MF_DI) ~ scale(DEM_DI) + I(scale(DEM_DI)^2) + scale(TPI_min_HR),
    data = occupancy.dataR)

Occupancy (logit-scale):
                                  Estimate    SE      z P(>|z|)
(Intercept)                    0.324 0.382  0.847 0.39699
scale(DEM_DI)             0.358 0.347  1.029 0.30333
I(scale(DEM_DI)^2)    -0.797 0.331 -2.411 0.01589
scale(TPI_min_HR)    -1.121 0.388 -2.891 0.00384

Detection (logit-scale):
                                                                 Estimate    SE      z  P(>|z|)
(Intercept)                                                  -3.014 0.193 -15.62 5.21e-55
scale(Herbaceous_cover_near)              -0.376 0.172  -2.19 2.84e-02
scale(Litter_cover_near)                           -0.391 0.145  -2.70 6.88e-03
scale(juliandate)                                        -0.279 0.111  -2.52 1.18e-02
scale(FA_HR)                                              -0.508 0.152  -3.34 8.34e-04
scale(Coniferall_cover)                             -0.715 0.243  -2.94 3.31e-03
scale(MF_DI)                                               0.284 0.110   2.58 9.78e-03

AIC: 823.5797
Number of sites: 162
optim convergence code: 0
optim iterations: 50
Bootstrap iterations: 0

>   confint(mo140, type = "det", level = 0.85)
                                                                                 0.075      0.925
p(Int)                                                              -3.2919891 -2.7364584
p(scale(Herbaceous_cover_near))           -0.6228475 -0.1289598
p(scale(Litter_cover_near))                        -0.5987767 -0.1825844
p(scale(juliandate))                                     -0.4385172 -0.1195762
p(scale(FA_HR))                                           -0.7263815 -0.2889471
p(scale(Coniferall_cover))                          -1.0651706 -0.3645129
p(scale(MF_DI))                                             0.1259582  0.4429540

>   confint(mo140, type = "state", level = 0.85)
                                                                        0.075      0.925
psi(Int)                                                     -0.2264812  0.8740046
psi(scale(DEM_DI))                                -0.1425272  0.8577829
psi(I(scale(DEM_DI)^2))                        -1.2734157 -0.3213967
psi(scale(TPI_min_HR))                        -1.6788972 -0.5627118



Thank you so much for your help in advance! Please let me know if I can provide any further information. 

Hailey



Marc Kery

unread,
Jul 29, 2023, 12:51:14 PM7/29/23
to unmarked
Dear Hailey,

all estimates are on the logit-link scale. You must wrap a call to plogis() around to get things on the probability scale.

E.g., do this:

plogis(confint(mo140, type = "state", level = 0.85))

Best regards  --- Marc


From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of haileyj...@gmail.com <haileyj...@gmail.com>
Sent: Saturday, July 29, 2023 18:37
To: unmarked <unma...@googlegroups.com>
Subject: [unmarked] Interpreting 85% CIs overlap with zero for psi intercept
 
--
*** Three hierarchical modeling email lists ***
(1) unmarked (this list): for questions specific to the R package unmarked
(2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
(3) HMecology: for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2021) and Schaub & Kéry (2022)
---
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/c144063c-03a2-4769-8528-3e5f67c3a255n%40googlegroups.com.

Marc Kery

unread,
Jul 30, 2023, 10:26:08 AM7/30/23
to unma...@googlegroups.com
Dear all,

I was temporarily confused when I wrote this last email 🙂 Of course, you must NOT inverse-link transform the coefficient or slope estimates of an occupancy model, but only the link-scale estimates of the intercepts. Thanks to Beni Schmidt for pointing out my mistake.

Best regards --- Marc

> Gesendet: Samstag, den 29.07.2023 um 18:51 Uhr
> Von: "Marc Kery" <marc...@vogelwarte.ch>
> An: unmarked <unma...@googlegroups.com>
> Betreff: Re: [unmarked] Interpreting 85% CIs overlap with zero for psi intercept

>
> --
> *** Three hierarchical modeling email lists ***
> (1) unmarked (this list): for questions specific to the R package unmarked
> (2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
> (3) HMecology: for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2021) and Schaub & Kéry (2022)
> ---
> 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.

haileyj...@gmail.com

unread,
Jul 30, 2023, 2:51:15 PM7/30/23
to unmarked
Hi Marc, 

Thank you so much for your help and the clarification!  

Best, 

Hailey
Reply all
Reply to author
Forward
0 new messages