Dealing with spatial autocorrelation using distsamp

[Saâd HANANE]

Dear all,

I am currently working with distance sampling (distsamp). Everything is going well, except for one issue: dealing with spatial autocorrelation (SAC). My Moran's I test shows a highly significant result (p-value < 0.0001).

Could you please advise if there is a way to account for SAC in distsamp? Ideally, my estimates should be adjusted to consider SAC. Additionally, how should SAC be handled when predicting the effects of covariates included in the best AICc model, particularly for producing figures?

Thank you

[Marc Kery]

Dear Saad,

you can't formally deal with SAC in unmarked, other than trying to "explain it away" by adding new, and more informative, covariates.

For modeling SAC, you'd normally have to go down the JAGS/NIMBLE way, or, actually, your best stab would be the R package spAbundance. This provides for hierarchical distance sampling with SAC.

Best regards  -- Marc

---

From: unma...@googlegroups.com <unma...@googlegroups.com> on behalf of Saâd HANANE <sdha...@gmail.com>
Sent: Wednesday, December 17, 2025 13:18
To: unmarked <unma...@googlegroups.com>
Subject: [unmarked] Dealing with spatial autocorrelation using distsamp

 

--
*** 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 visit https://groups.google.com/d/msgid/unmarked/424b25e8-0e99-4fec-bb65-adf4badcf908n%40googlegroups.com.

[Jeffrey Royle]

Dear Saad,

  In addition to what Marc said (with regard to how to do it), I would suggest that SAC is pretty unimportant in general as long as your study design is not too pathological. For example if you did a random or systematic, or systematic random sample then the existence of unaccounted for spatial structure (that would produce SAC) probably only induces a little bias in the variance estimates. You could adjust for this using a lack-of-fit or quasi-likelihood adjustment (how poor is the model fit, anyways?).  It might be better to try and identify the relevant source of the SAC (i.e., missing habitat information) and model it that way but if you feel it's really important then spAbundance (or JAGS/NIMBLE) is the way to go.

 As always it would be interesting to see a simulation study done to characterize exactly how bad you do (and in what manner) by not accounting for the SAC.

regards

andy

--