bootsum.p: Summarise detection probability bootstrap results.

In DistanceDevelopment/hsltm: Line transect estimation with hidden state availability models

Description

Uses bootstrap results from bootstrap.p.with.Et or bootstrap.p.with.hmm to work out bootstrap means, variance estimates, CVs and confidence intervals.

Usage

bootsum.p(bs, probs = c(0.025, 0.975), pcut = 0)

Arguments

bs	output from bootstrap.p.with.Et or bootstrap.p.with.hmm.
probs	lower and upper percentile points for confidence interval reporting.
pcut	minimum estimated detection probability to use. This is a quick and dirty method to robustify against small detection probability estimates skewing the distribution of 1/phat badly for small samples. It is ad-hoc. If you use it, do histogram of $bs$phat to see if there is a reasonable cutpoint.

Value

Returns a list with elements

• nboot: number of bootstrap estimates used in constructing bootstrap statistics.

• nbad: number of bad estimates excluded from results.

• parcov: parameter estimate variance-covariance matrix.

• parcorr: parameter estimate correlation matrix.

• bests: bootstrap estimate statistics, comprising meand, standard error, percentage CV and confidence interval limits for: (a) estimated mean detection probability, (a) 1/(estimated mean detection probability), (c) estimated p(0), ad (d) detection hazard function parameters.

DistanceDevelopment/hsltm documentation built on June 21, 2019, 2:22 p.m.