Hi all,
I am working with the multi-species models developed by Rota et al (2016) to investigate co-occurrence patterns of three species: the bobcat, gray fox and cottontail rabbit. I fitted several models with different hypotheses of species co-occurrence and chose the best models with the AICc. My best model indicates that there is a "spatial interaction" between two species pairs Grayfox vs Cottontail and Bobcat vs Cottontail.
My model looks like this
m8c <- occuMulti(c("~Vertcover_50", "~Effort", "~Cam"), c("~Neob", "~1", "~Neob",0 , "~Neob", "~1" ,0), data=multi)
Where Neob is a two-level factor: 1 for vegetation dominated by cacti and 0 for other vegetation.
I am interested in calculating the species interaction factor (SIF) or odds ratio v (MacKenzie et al 2018, p.530). SIF=(ψ_11/(1-ψ_11))/(ψ_10/(1-ψ_10))
Since it is a derived parameter and to get the confidence intervals, it must be estimated by the delta method or by parboot. I am trying to do it by parboot using the following function:
SIF_par
<- function(fm) {
Psi11f <-
plogis(coef(fm)[["psi([gray_fox:rabbit] (Intercept))"]])
minPsi11f <-
1-Psi11f
Psi10f <-
plogis(coef(fm)[["psi([gray_fox] (Intercept))"]])
minPsi10f <- 1-Psi10f
Psi11b <-
plogis(coef(fm)[["psi([bob_cat:rabbit] (Intercept))"]])
minPsi11b <- 1-Psi11b
Psi10b <-
plogis(coef(fm)[["psi([bob_cat] (Intercept))"]])
minPsi10b <- 1-Psi10b
SIFf
<- (Psi11f/minPsi11f)/(Psi10f/minPsi10f)
SIFfb <- (Psi11b/minPsi11b)/(Psi10b/
minPsi10b)
total
<- c(siff=SIFf, sifb=SIFfb)
return(total)
}
trySIF
<- parboot(m8c, SIF_par, nsim= 1000)
SIFCI <-
cbind(trySIF@t0, t(apply(try...@t.star,
2, quantile, probs= c(0.025, 0.975))))
I manage to get the estimates and intervals, but the values are huge. For example, for the fox and rabbit the SIF= 462 with 95%CI = 9.56 --1170482927, and for the bobcat and rabbit it is similar but sadder SIF= 10.50108; 95%CI= 0.22 -- 12071948.
Is something wrong with my function? Should I try the delta method? Is it possible to calculate the SIF in relation to my categorical covariate? Or is the function OK and my data is definitely not very good?
Thanks in advance for the help
--
Gabriel P. Andrade-Ponce
Biólogo - Universidad Nacional de Colombia
MSc y Estudiante de Doctorado -Instituto de Ecología A.C.
Xalapa, Veracruz, México