Empirical likelihood ratio tests and confidence intervals for means/quantiles/hazards from possibly censored and/or truncated data. In particular, the empirical likelihood for the Kaplan-Meier/Nelson-Aalen estimator. Now does AFT regression too.
This package is based on R code el.cen.EM, el.cen.EM2,
el.trun.test emplikH1, emplikH2, emplikdisc.test and
el.cen.test, etc. written by Mai Zhou.
The code el.test is ported from S code ELM() by Art Owen.
el.test() can handle vector means as null hypothesis (several
mean parameters), but data must be uncensored.
When dealing with censored data, use el.cen.EM2() to handle
vector means.
Many other code in this package can only handle one parameter in H0.
But that parameter(s) can be mean/hazard/quantile, and the
function can also take censored/truncated data as inputs.
The name convention: el.xxx for the parameter of mean(s).
emplikxxx for the parameter of hazard(s).
Release History:
Pre-beta 1998--1999
Version 0.1-2 Nov. 12, 2001
Version 0.2-1 Jan. 10, 2002
Version 0.3 Feb. 18, 2002
Version 0.4 Apr. 6, 2002
Version 0.5 Apr. 24, 2002
Version 0.5-1 Apr. 25, 2002
Version 0.6 May 13, 2002
Version 0.7 Oct. 12, 2002
Version 0.7-3 Jan. 12, 2003
Version 0.7-4 July 20, 2003
Version 0.8 March 18, 2004
Version 0.8-1 June 14, 2004
Version 0.8-2 Sept. 10, 2004
Version 0.8-3 March 24, 2005
Version 0.8-4 March 25, 2005
Version 0.9 April 15, 2005
Version 0.9-1 April 24, 2005
Version 0.9-2 Oct. 9, 2005
Version 0.9-3 Aug. 2006
Add function findLnew(), findUnew(). 6/2022.
Add function el.cen.kmc1d() in version 1.1-2, which is much faster
than el.cen.EM(). But it sometimes fail to find the right answer.
Add el.cen.test2() in version 1.3-3.
To do list:
Speed up with solve(A,b) instead of finding the
full inverse matrix by svd and compute A{-1}*b .
See emplikH2.test2() for an example.
Integrate the kmc package into emplik package, may be.