Robust and alternative estimators for "better" estimates for expenditures and other "long tail" distributions.

Abstract A 2006 Tourism Management article proposes using specific robust estimators to determine “better” estimated means for long-tail distributions; that is for skewed distributions with valid large responses heavily influencing the mean. Getting better estimates matters because long-tail distributions occur frequently for amounts and quantities. In addition, long-tail distribution sample means and totals can be so variable using those prompts concerns. However, low variability robust estimates of means and totals can be badly biased. Therefore, a focus of this paper is obtaining relatively low variability estimates that are not “too” biased. Real data are used to illustrate attributes of long-tail distributions. Results show some robust estimators suggested for producing better estimates are badly biased and therefore not better. Three ways of obtaining lower variability estimated means and totals that are not “too” biased are discussed. Practical and research implications of the ideas presented and of results obtained are discussed.