A REVIEW OF SKEWNESS AND KURTOSIS MEASURES (incomplete)
DESCRIPTION
This is an extended review and analysis of skewness and kurtosis measures in statistics. It starts with Karl Pearson's papers in 1895 which defined the concepts. It covers Fishers contributions and some of the use of these measures by early scientists. The historical view was to use these measures to find appropriate statistical distributions that could be used on a specific data set. The equations that were used are included. Current views are generally that of using the normal distribution, and to use skewness and kurtosis measures as a statistic. However they are still used as measures of normality/non-normaiity of data sets.
Right now, this paper is being rewritten. Only sections I and 2 are close to completion. Work has been stopped, because of other projects, and a lack of interest in the issues.
Email me at 'd_heiser@att.net' or 'daheiser594@gmail.com' if you have any comments or questions.
CONTENTS
1. INTRODUCTION
2. HISTORIC DEVELOPMENTS
3. UNIVARIATE MEASURES AND TESTS
3.1 SKEWNESS
3.2 KURTOSIS
3.3 COMBINATIONS OF SKEWNESS AND KURTOSIS
3.4 SYMMETRY AND TAILS
3.5 COMPUTATION METHODS
4. MULTIVARIATE MEASURES AND TESTS
4.1 SKEWNESS
4.2 KURTOSIS
4.3 MATRIX REDUCTION METHODS
5. REFERENCES
1. INTRODUCTION
2. HISTORICAL DEVELOPMENTS
ACKNOWLEDGEMENTS
I would like to thank Mike Palij, New York University for providing *.pdf copies of some of the old Pearson and Pearl papers from JSTOR.
COMMENTS AND REMARKS
This effort started with a paper for the Sacramento Statistical Association's Institute for the year 2000. The version here is an expanded and corrected one from the old paper.
CORRECTIONS, CHANGES, ADDITIONS AND NOTES