On Thursday, October 18, 2012 2:40:45 AM UTC-5, Ray Koopman wrote: > On Oct 13, 3:45 pm, Paul <paulvonhip...@yahoo.com> wrote: > > > Bias is important. The quantity that I am estimating, gamma, is a variance that will be used to calculate confidence intervals. If the estimate of gamma is negatively biased, then the coverage of the confidence intervals will be too low. > > > > > > What is an appropriate loss function to use under these circumstances? > > > > One easy fix would be to add a penalty function, say some multiple of > > log(1-gamma). Then all that remains is to justify it.
Thanks, this is interesting. Are you suggesting a penalized likelihood, or would the penalty be used in another way?