icenReg (version 1.3.5)

ICNPMLE: Computes the NPMLE for Univariate or Bivariate Interval Censored Data

Description

Computes the MLE for a Interval Censored Data with a squeezing EM algorithm (much faster than the standard EM). Accepts either univariate interval censored data (where times is an n x 2 matrix with times[,1] being the left side of the interval and times[,2] is the right side), or bivariate interval censored data (where times is an n x 4 matrix with times[,1:2] being left and right side of the interval for event 1 and times[,3:4] being the left and right side of the interval for event 2).

Usage

ICNPMLE(times, B = c(1,1), max.inner = 100, max.outer = 100, tol = 1e-10)

Arguments

times

either an n x 2 or n x 4 data.frame or matrix of censoring intervals

B

A vector indicating whether each end of the intervals are open (0) or closed (1). Alternatively, this could be an n x 2 or n x 4 matrix of indicators for each individual interval

max.inner

number of inner loops used in optimization algorithm

max.outer

number of outer loops used in optimization algorithm

tol

numerical tolerance

References

Anderson-Bergman, C., (2014) Semi- and non-parametric methods for interval censored data with shape constraints, Ph.D. Thesis

Yu, Y., (2010), Improved EM for Mixture Proportions with Applications to Nonparametric ML Estimation for Censored Data, preprint

Maathuis, M., (2005). Reduction algorithm for the NPMLE for the distribution function of bivariate interval censored data. Journal of Computational and Graphical Statistics Vol 14 pp 252- 262

Examples

simData <- simBVCen(500)

fit <- ICNPMLE(simData)

fit