Statistical inference for a linear function of medians: A comparison of the Maritz-Jarrett and Price-Bonett estimators
Researchers in the behavioral and social sciences usually compare independent groups by analyzing means. However, when one investigates real-world data, it becomes evident that the majority of the underlying distributions are either highly skewed or heavy tailed. This fact suggests that an analysis of medians may be a more appropriate and meaningful way to compare independent groups. An analysis of medians procedure, derived by Price and Bonett (2002), claimed to generate the most robust, efficient, and accurate estimators of location; particularly when the sizes of the independent groups being compared were small. In order to investigate this claim, Monte Carlo techniques were employed to construct confidence intervals around the sample median and the Harrell-Davis estimator using the Maritz-Jarrett and Price-Bonett standard error estimates, respectively. Sample observations were randomly generated from three mathematical and four empirical distributions in order to compare the confidence intervals for a single population median, the difference in two population medians, pair-wise differences in medians, as well as the simple, main, and interaction effects of a 2X2 factorial design. The results of the study showed that the Price-Bonett estimator tended to outperform the Maritz-Jarrett estimator when the average coverage probability and the confidence interval widths were used as indicators.
|Year of publication:||
|Authors:||Lawson, Kevin Duwan|
Wayne State University
|Type of publication:||Other|
ETD Collection for Wayne State University
Saved in favorites
Similar items by subject
Find similar items by using search terms and synonyms from our Thesaurus for Economics (STW).