bandwidth.choose: Selection of optiaml bandwidth for nonparametric density...

In IsoCI: Confidence intervals for current status data based on transformations and bootstrap.

Description

Conduct likelihood-based leave-one-out croos-validation (loocv) to select optimal bandwidth for nonparametric density estimation of currenst status failure times.

Usage

bandwidth.choose(h.set, z, d)

Arguments

h.set	a set of bandwith values
z	a vector of covariate
d	a vector of outcome

Value

h.opt	an optimal bandwidth
result.table	the 1st column: bandwidth, the 2nd column: loocv-log-likelihood value

Author(s)

Choi, B. Y., Fine, J. P., and Brookhart, M. A.

References

Choi, B. Y., Fine, J. P., and Brookhart, M. A. (2013) Practicable confidence intervals for current status data. Statistics in Medicine 32, 1419-1428.

Ghosh, D., Banerjee, M., and Biswas, P. (2008). Inference for Constrained Estimation of Tumor Size Distributions. Biometrics 64, 1009-1017.

Groeneboom, P. and Wellner, J. A. (1992). Information Bounds and Nonparametric Maximum Likelihood Estimation. Boston: Birkhauser.

See Also

iso.ci

Examples

# simulating data

n <- 50
z <- rexp(n)
pz <- pexp(z)
d <- rbinom(n,1,pz)

# finding optimal bandwidth for estimationg a density function
h.opt = bandwidth.choose(h.set=seq(0.1,2,len=15),z=z,d=d)

# Untransforemd and transformed Wald-type confidence intervals
fit.wald <- iso.ci(z=z,d=d,h.opt=h.opt$h.opt)

# Bootstrap confidence intervals
## Not run: fit.bt <- iso.ci(z=z,d=d,method="bt",nboots=100) 

# Untransforemd and transformed bootstrap-Wald-type confidence intervals
## Not run: fit.bt.wald <- iso.ci(z=z,d=d,method="bt.wald",nboots=100)

IsoCI documentation built on May 2, 2019, 8:31 a.m.