How it works

Overview

AbundanceR generates species and group specific abundance estimates using multi-model inference and Huggins removal models in 
program MARK. Data analysis is a five step process: 1) the user loads data from the USGS point count database, 2) upon import, AbundanceR truncates the data set at the last finite distance band (e.g. 100m) and removes flyovers, 3) the user selects target species and assigns target sites to up to three different groups for comparison, 4) the program sends input files representing multiple models to program MARK for analysis, and 5) the program derives model-averaged estimates of abundance and detectability based on output from program MARK.

Models considered for analysis

When selecting only a single species, AbundanceR runs a null detection model {p(.)} and a group-specific detection model {p(group)} in program MARK. When selecting multiple species, AbundanceR runs a null detection model {p(.)}, a group-specific detection model {p(group)}, a species-specific detection model {p(species)}, and a detection model with a group by species interaction {p(group*species)} in program MARK.

Model convergence

When there is not enough information for program MARK to estimate species-specific detectability, as indicated by extremely high standard errors or standard errors of zero, AbundanceR removes species with non converging estimates and reanalyzes all candidate models. This process continues until there is enough information to estimate species-specific detectability for all species.

To more easily detect non-convergence in models estimates, abundanceR uses the logit link instead of the default sin link to generate estimates.

Model-averaged estimates

AbundanceR uses the formulas presented in 
Chapter 14 of Program Mark: A Gentle Introduction to calculate model-averaged estimates of abundance. The program then generates model-averaged estimates of detectability, the probability of detecting an individual during the entire count, as the number of detected individuals divided by the predicted abundance. Note that the detectability estimates typically produced by program MARK are not count-specific but interval-specific. To convert the interval- specific detectability estimates typically produced by MARK to the count-specific detectability estimates produced by AbundanceR use the following equation:
                
p = 1-(1-p*)^5

Where p is the detectability estimate from AbundanceR and p* is the detectability estimate from MARK.
Comments