Software for Computing Empirical Likelihood Ratio

For uncensored data, when the parameter is a mean vector, Art Owen wrote an Splus function that computes the empirical likelihood ratio. I have port it to R. (included in the R package emplik, see below).
For right, left or double censored data, and for left truncated, right censored data, the functions in the R package emplik can handle the computation of empirical likelihood ratio. In fact, the parameters being tested there can be either expressed as the `weighted, integrated hazard', or weighted mean. Of course, those can be used when data are not censored.

The current version of emplik is 0.8 released on March 18, 2004. The latest version should be available in CRAN. Check out http://cran.us.r-project.org Look in the contributed package area.
I have wrote two R functions. One for discrete hazard, one for continuous hazard. Of course, these can be used for uncensored data too.
For right censored data, when parameter is (weighted) mean, the computation is a little involved. I have written two types of functions for that case. One uses sequential quadratic programming (SQP) computation to find the maximum. It may be slow/memory hungary for larger samples. But for sample size up to 2000, 30% right censoring you can get the result under 1 minute on a 1 GHz 256MB machine. The other uses EM algorithm, called el.cen.EM etc. Both need iteration. For sample size up to 10000 this one works OK even on a laptop computer. (1 minute)
For doubly censored observations (some data are right censored, some data are left censored), I also have written a few functions for that purpose. One uses sequential quadratic programming (it may be slow/memory hungary for larger samples), with weighted mean as parameter. One uses EM algorithm with parameter of F(T), probability at a give time T. The other also uses EM algorithm but the parameter is the mean. All 3 need iteration.

coxEL R package/function for Cox model with extra info on baseline. This is alpha 1.0 release.