Citation

Yearout R, Barger R, Yates G, Lisnerski D. Int. J. Ind. Ergonomics 1999; 24(1): 129-134.

Copyright

(Copyright © 1999, Elsevier Publishing)

DOI

unavailable

PMID

unavailable

Abstract

Data collected by practicing industrial ergonomists may not conform to the assumptions of normality; i.e., that error terms are independently and identically distributed. For normality, error terms within groups also must have a mean equal to zero and equal variances ([delta]2). It is not unusual for both of these assumptions to be violated. When this heterogeneous condition occurs, the results of commonly used parametric techniques can be inappropriate. Strengths and weaknesses of several statistical tests for homogeneity of [delta]2 and their appropriateness are discussed. The Levene's Test, with computer application, is used to demonstrate its application for typical data. Both parametric and non-parametric tests are used to illustrate that significant differences may be inidicated when in fact there are none. This paper is intended to be a resource for the ergonomics practitioner in determining initial data analysis procedures.Relevance to industryThe ergonomics practitioner and engineer realizes that industrial data collected in field studies cannot be controlled like a classical laboratory study. In this environment much of the data obtained contains both unequal sub-group sample sizes and variances. Thus traditional parametric statistics are inappropriate for analysis and inferences. This paper provides industrial practitioners and engineers with the tools and guidelines that are appropriate for conducting statistical analysis of typical field data, drawing inferences, and making meaningful conclusions.