Stochastic Efficiency, Normality, and Sampling Errors in Agricultural Risk Analysis

Abstract

Abstract This paper examines the power of tests for efficiency. The relationship between sample size, parameter values, and the family of probability distributions is stressed. Some findings are that the probability of correctly ranking distributions is frequently very low regardless of sample size. It is generally lowest as distributional parameters (such as the means) of the two distributions being compared are of similar magnitudes. Further, the empirical distribution function performs extremely well as compared to maximum likelihood estimators.