Robust Estimation of Location and Scale Parameters.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OHIO SCHOOL OF ENGINEERING
Pagination or Media Count:
A Monte Carlo analysis was performed to determine the efficiency of the Harter robust estimators of the scale and location parameters of double exponential, normal, and uniform distributions relative to the maximum-likelihood estimators. Two modifications were made to the Harter estimators which increased the relative efficiency except when the underlying population was uniformly distributed. The modified estimators were then designated the Moore estimators and the Jorgenson estimators. Tables were prepared comparing the relative efficiencies of the Harter, Jorgenson, Moore, Hogg, Hodges-Lehmann, and the Switzer robust estimators of the location parameter for the double exponential, the normal, and the uniform distributions for samples of size 12 and 24. Two types of figure of merit were defined for a robust estimator. The choice of the best robust estimator is a function of sample size and the criteria used. Author
- Statistics and Probability