I can not find any documentation on the type="bca" for the R command confint used to find confidence intervals. I am using this function to find the confidence interval of a bootstrap model. I know there are normal CI, percentile, and pivotal. Does the bca correspond to pivotal or something else?
2 Answers
From ?boot::boot.ci
:
bca: The intervals calculated using the adjusted bootstrap percentile (BCa) method.
See references therein, particularly Davison and Hinkley.
The code that actually does this is in the (hidden) function boot:::bca.ci
. The comments in the source code (here say:
#
# Adjusted Percentile (BCa) Confidence interval method. This method
# uses quantities calculated from the empirical influence values to
# improve on the precentile interval. Usually the required order
# statistics for this method will not be integers and so norm.inter
# is used to find them.
#
BCa is the bias-corrected with acceleration constant method for confidence interval estimation. See my book Bootstrap Methods: A Guide for Practitioners and Researchers, 2nd Edition, Chernick, 2007, or any other major text on bootstrap such as An Introduction to the Bootstrap, by Efron and Tibshirani, 1994.