Calculate Jaccard index from a Multi Species Occupancy Model
Posted: Thu Sep 11, 2025 3:41 pmHello,
I'm working with multi-species occupancy models in RPresence and have fit models with species-specific detection covariates for insects and birds. This allowed me to estimate species richness for my sites after accounting for imperfect detection. The estimated number of species from my best model increased for insects quite a bit (the most speciose site went from 128 species to 162 species) but stayed relatively similar to our sampled data for birds. The R^2 between the sampled richness and msom-estimated richness is 0.998 for insects and 0.996, so I think this implies that our original sampling effort is representive, but I've continued our analysis with the estimated species richness values to better estimate effect sizes.
I'm looking to extend these models to our community-level analyses and use Jaccard's index to measure similarities between sites, but I am unsure how to add the 'undetected' species to my presence/absence matrix. The conditional occupancy (psi) for most species is very low (<0.10), with only a few species reaching psi_c values of 0.20-0.25, so I'm not sure I can confidently say that a species was there and simply went undetected. I've seen this type of implementation done in JAGS and other software but it's quite technical and a bit tough to follow. I am wondering if there is a solution I can implement in RPResence or if anyone has advice on how to predict the species presence/absence matrix that incorporates these occupancy estimates.
Thanks.
I'm working with multi-species occupancy models in RPresence and have fit models with species-specific detection covariates for insects and birds. This allowed me to estimate species richness for my sites after accounting for imperfect detection. The estimated number of species from my best model increased for insects quite a bit (the most speciose site went from 128 species to 162 species) but stayed relatively similar to our sampled data for birds. The R^2 between the sampled richness and msom-estimated richness is 0.998 for insects and 0.996, so I think this implies that our original sampling effort is representive, but I've continued our analysis with the estimated species richness values to better estimate effect sizes.
I'm looking to extend these models to our community-level analyses and use Jaccard's index to measure similarities between sites, but I am unsure how to add the 'undetected' species to my presence/absence matrix. The conditional occupancy (psi) for most species is very low (<0.10), with only a few species reaching psi_c values of 0.20-0.25, so I'm not sure I can confidently say that a species was there and simply went undetected. I've seen this type of implementation done in JAGS and other software but it's quite technical and a bit tough to follow. I am wondering if there is a solution I can implement in RPResence or if anyone has advice on how to predict the species presence/absence matrix that incorporates these occupancy estimates.
Thanks.