calibSI: Attempts to find a support interval with the desired...

In supportInt: Calculates Likelihood Support Intervals for Common Data Types

Description

This function performs a grid search to find the lowest support level that has the desired coverage probability. It begins by searching a grid of relative likelihood values using a fixed number of parametric bootstraps at each point. The likelihood value is then refined using a golden section based search.

Usage

calibSI(dat, n = NULL, family, conf.level = 0.95, B = 2000, 
gridlo = 4, gridhi = 20, gridix = 2, tol = 0.03)

Arguments

dat	This argument is the number of success if family='binomial', the number of occurences if family='poisson', or a vector of normal draws if family='normal'.
n	Number of trials if family='binomial'
family	Data type entered, accepts "binomial", "poisson", "pois", "gaussian", "normal"
conf.level	target confidence level
B	Number of parametric bootstraps to use to estimate coverage probability after the initial grid search.
gridlo	Lowest relative likelihood considered in the initial grid search
gridhi	Highest relative likelihood considered in the initial grid search
gridix	size of steps to be checked between gridlo and gridhi in initial search
tol	acceptable error in the likelihood support level returned

Value

si	The support interval estimated to have the desired confidence level
support.level	The relative likelihood defining the interval's endpoints
init.grid	Relative likelihoods checked as part of the initial gridsearch and their estimated coverage

Author(s)

Derek Smith

Examples

set.seed(10)
calibSI(rnorm(15), family="normal", conf.level=.95, B=1000)

supportInt documentation built on May 2, 2019, 6:32 a.m.