A report on the comparison of extreme quantiles estimators in surveys.
Hua Liu, M.S. Candidate, Department of Statistics, Colorado State University.
Friday, March 27, 2009
3:00 pm, Statistics 006
In the real world, extremal events happen almost everywhere and many heavy tail estimators have been developed. In this report, we compare the practical performance of several common extreme value estimators when applied to the estimation of extreme quantiles in a finite population. We compare the Hill estimator and the Dekkers-Einmanhl-deHaan (DEH) estimator against a traditional Horvitz-Thompson (HT) estimator. Through a simulation study, we obtain the biases and squared deviations generated from Hill, DEH and HT estimators in different scenarios (different population distribution, sample sizes, parameters, sampling designs, etc), and discuss the performance of each estimator..
Advisor: Jean Opsomer
Member: Jay Breidt
Outside: Hong Miao(Finance and Real Estate)