Sample Size issues with orangutan nest survey

[Beth Barrow]

Hello,

We completed a survey of orangutans in a National Park with a mosaic of habitats covering both primary and secondary growth forest types.  In total, we surveyed 50 lines each 1km in length. 

However, when accounting for these different habitats and the different nest degradation multipliers that come with that (I have 5 different degradation values and 7 habitat groupings), the sample size is reduced to way below the minimum number of lines needed for robust analysis.  When grouping transects based on different habitats/forest types the # samples varies from just 3 (montane forest) to 17 (secondary growth mixed lowland)

I am wondering if you can offer advice as to the best way forward with such a dataset?  Should we avoid attempting to estimate a population size completely due to the small sample sizes? Or is it still worth trying to do something with these data?

Many thanks

[Eric Rexstad]

Beth

It is common to have small numbers of spatial replicates in strata for which you wish to make inference.  But you do have some options available to you.

The most simple (from an analysis perspective) is to reduce the number of habitats/degradation classes; trying to combine those that are similar into a sufficiently small number of categories with sufficient replicates in each category.

More analytically complex is to employ habitat classes as strata and use the strata as covariates in the detection function.  This way, stratum-specific detection functions are estimated using a key function that is shared across all strata, but the scale parameter of the key function differs for each stratum.

I have an example of this second type of analysis using species rather than habitat type as strata, with estimation at the species level for some species for which the number of detections is small.  This example differs from yours in that the number of spatial replicates is high for all species. 

http://examples.distancesampling.org/Distance-spec-covar/species-covariate-distill.html

Covariate modeling with rare species - Distance Sampling Species-specific density estimates. Density estimates for each species can be produced by using the dht2 function that contains the argument strat_formula used to specific the levels of stratum-specific estimates requested. The stratification argument ensures the correct measures of precision are associated with the species-specific density estimates. The value object indicates this analysis ... examples.distancesampling.org

In your situation, the covariate approach may help you with detection function modelling, but it does not address the low level of spatial replication in some strata.  In strata with few replicate transects, the encounter rate variance between transects will be poorly estimated and likely will lead to poor precision in the abundance estimates for those strata.  There is faint hope that density surface modelling might help with the low number of spatial replicates, but I'm not certain of this.

---

From: distance...@googlegroups.com <distance...@googlegroups.com> on behalf of Beth Barrow <bethj...@gmail.com>
Sent: 04 November 2021 14:34
To: distance-sampling <distance...@googlegroups.com>
Subject: [distance-sampling] Sample Size issues with orangutan nest survey

 

--
You received this message because you are subscribed to the Google Groups "distance-sampling" group.
To unsubscribe from this group and stop receiving emails from it, send an email to distance-sampl...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/distance-sampling/ed7b0018-85bf-40e6-bfd8-1ed9fe9d0f19n%40googlegroups.com.