UBC Theses and Dissertations
Documenting the impact of outliers on Cronbach’s coefficient alpha estimate of reliability : informing how one should interpret the extant literature and/or one’s own research findings Liu, Yan
The study of outliers and their impact on statistical results is important because outliers are prevalent in real data. Previous research has focused on outlier identification and treatment with no studies having investigated whether outliers have impact on the estimated Cronbach's coefficient alpha. This is an important gap in the research literature because coefficient alpha is the most widely used measurement statistic in all of the social sciences. The purpose of this study is to investigate whether outliers impact the value of coefficient alpha at different amounts of outlier contamination, the distance of the center of the contamination population distribution from the uncontaminated parent population, and standard deviation of the contamination population for varying values of population reliability and sample sizes. The overall results indicate that the estimates of coefficient alpha are not affected by symmetric outlier contamination (i.e., when the means of the contamination and parent distributions are equal), whereas asymmetric outlier contamination artificially inflates the estimates of coefficient alpha. In particular, this upward bias of coefficient alpha estimates and inflation in statistical efficiency (i.e., stability) are greater with increasing asymmetry and proportion of outlier contamination. These effects of outliers on the bias and efficiency of coefficient alpha estimates are reduced for increasing population reliability.
Item Citations and Data