Nonparametric Bootstrap For Year-stratified N-mixture Model

[Tyler Hodges]

Hello group,

I have a question regarding the best way to implement a nonparametric bootstrap on a year-stratified N-mixture model. I am using a large dataset of point-counts that spans three years. Sites were generally surveyed twice in any given year. However, some sites were surveyed for two years, and others for only one (this is a subset of a larger point-count dataset, so for some sites, either the first year or second year of surveys are being excluded). In total, there are 876 unique points, and 1341 "sites" after stacking those that were surveyed for two years. 

I was wondering how to set up a nonparametric bootstrap for a dataset with this much variability in number of years each site was surveyed? Does/can the example bootstrap outlined in section 2.3.4 of AHM v2 account for variability in number of years each site was surveyed?

And, a related question: should year always be included as a fixed effect on abundance in all models to help account for the breaking of the temporal structure and any "noise" that may introduce? 

Thanks in advance for your help!

Best,

Tyler

[Marc Kery]

Dear Tyler,

 

I think I’d resample sites (i.e., with all years of data for a chosen site). Then, now that unmarked can do some extra random effects when choosing engine = ‘TMB’, I’d probably not stack, but fit the model with site-level random effects in abundance.

 

Best regards  ---- Marc

--
*** 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/b92f221e-f45f-4e65-807c-7f4f0c6976b7n%40googlegroups.com.

[Tyler Hodges]

Thanks Marc! 

I was unaware that unmarked now supports random effects! In that case, if I still wanted to utilize the stacked structure, would rerunning all of my models with a random effect on abundance of either the point i.d. or year (rather than just a fixed effect of year) be the best way address the pseudoreplication inherent in the stacked structure? I briefly tried to do just that with either (1|Year) and (1|Point) on both my top model (3 obs covs and 3 site covs) and a null model with no covs and each time ran into convergence issues without providing starting values. I wonder if, despite my rather large dataset, it may still not be large enough to include random effects that are made up of categorical variables (in the case of year, it would be a factor with 876 levels!)? 

Thanks again!

Tyler

[Marc Kery]

Dear Tyler,

I'd only fit random site effects, but not also random effects for the 3 (and not 876) levels of the factor year. Although it would not be technically wrong to fit also random year effects, in practice doing that with only 3 levels for a factor may lead to numerical grief.

Best regards  --- Marc

---

From: 'Tyler Hodges' via unmarked <unma...@googlegroups.com>
Sent: Tuesday, December 5, 2023 21:30
To: unmarked <unma...@googlegroups.com>
Subject: Re: [unmarked] Nonparametric Bootstrap For Year-stratified N-mixture Model

 

To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/825696cc-a36a-4db7-b4b7-bb0be74e88dcn%40googlegroups.com.

[Tyler Hodges]

Hello Marc,

Haha, nice catch! I meant that point i.d./site would have 876 levels. 

Best,

Tyler