Additional example analyses
Posted 10 March 2020 by Eric Rexstad
Two situations where covariates provide the best solution to difficult analytical situations
Additions to case study library
The library of example analyses available online at examples.distancesampling.org has grown to 14. The most recent additions address situations in which I believe covariates provide the best solution to difficult analytical situations. Those situations are
multi-species or multiplie geographic strata where species/stratum-specific abundance estimates are required, but some species or strata have a small number of detections. In this situation, pooling data across species or strata might be used to estimate the detection function, but the pooled detection function should not be used to make inference to the individuals species or stratum. This is a situation where species/stratum as a factor covariate can be used to estimate an appropriate species/stratum-specific detection function by borrowing strength for parameter estimation from other species/strata. The example used is Buckland’s (2006) Montrave songbird survey. The vignette is located here.
correction for size bias. Bias is the estimation of average group size in the population is induced in distance sampling surveys because small groups at large distances are less likely to be detected and included in the sample. This positive bias then biases the estimates of individual abundance because abundance of individuals is derived as the product of number of groups and average group size. Recognising that probability of detecting groups is a function not only of distance from the transect, but also size of the group leads to incorporation of group size as a covariate in the detection function. In this fashion, positive size bias in estimated group size is eliminated. Example analysis is of spotted dolphin pods surveyed from tuna fishing vessels. Read the vignette here.