Jaatha.confidenceIntervals: Function for calculating bootstrap confidence intervals for jaatha estimates.

Description

This functions calculates bias-corrected and accelerated (BCa) bootstrap confidence intervals as described in Section 14.3 of "An introduction to the bootstrap" by Efron and Tibshirani. Basically, we simulate many datasets under the model using the estimated parameters and do a complete Jaatha estimation procedure on each dataset. We can then use the thereby estimated values to calculate approximate confidence intervals.

Usage

Jaatha.confidenceIntervals(jaatha, conf.level = 0.95, replicas = 100,
  cores = 1, log.folder = tempfile("jaatha-logs"), subset = 1:replicas)

Arguments

jaatha

A jaatha object that was returned by Jaatha.refinedSearch.

conf.level

The intended confidence level for the interval.

replicas

The number of Jaatha runs that we perform for calculating the confidence interval. Should be reasonable large.

cores

The number of CPU cores that will be used.

log.folder

A folder were log-files for the indiviual runs are placed. The default is to use a temporary folder, even though it is highly recommended to specify a folder and save the logs.

subset

This setting allows you to distribute the CI calculation on multiple machines. You can specify which replicas should be run on this machine. Each run is identified by an integer between one and replicas. All runs which integers are passed as a vector in t

Value

The Jaatha Object with confidence intervals included if 'subset' was not used. Nothing otherwise.

Details

Warning: This requires a large number of Jaatha runs and should be best executed on a small cluster rather than on a desktop computer.