Colloquium The Department of Statistics presents   Nicole Lazar Department of Statistics,
Carnegie Mellon University     SOME INFERENTIAL ASPECTS OF EMPIRICAL LIKELIHOOD     Abstract Empirical likelihood was proposed as a non-parametric analogue of ordinary likelihood. It is known that for inference regarding smooth functions of the mean, the high order asymptotic properties of ordinary likelihood, such as Bartlett correctability, are inherited by empirical likelihood. We show that in situations where nuisance parameters are incorporated into the problem via a system of estimating equations, these higher order properties break down.

The efficiency of empirical likelihood is also explored. We find that to second order, empirical likelihood has the same power against contiguous alternatives as does an ordinary parametric likelihood, which is taken to represent the true state of nature. To third order, there is a difference in power, which may be in favor of either of the models.

Finally, we compare empirical likelihood and quasi-likelihood in terms of their conditional properties. Implications of the various results will be discussed.     Monday, March 23, 1998 3:00 P.M.,  1064 CEB (Duncan Hall)  

 
 

back to Department of Statistics Home Page This page last maintained on 03/12/98 Problems or suggestions to webmaster@stat.rice.edu Contact the Department of Statistics via stat@stat.rice.edu