Date of Award
Spring 5-2022
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
First Advisor
Dr. Marita Apter-Desselles
Abstract
Recent research suggests that individual productivity may not be normally distributed and is best modeled by a power law, a form of a heavy-tailed distribution where extreme cases on the right side of the distribution affect the mean and skew the probability distribution. These extreme cases, commonly referred to as “star performers” or “productivity stars,” provide a disproportionately positive impact on organizations. Yet, the field of industrial-organizational psychology has failed to uncover effective techniques to identify them during selection accurately. Limiting factors in the identification of star performers are the traditional methods (e.g., Pearson correlation, ordinary least squares regression) used to establish criterion-related validity and inform selection battery design (i.e., determine which assessments should be retained and how those assessments should be weighted). Pearson correlation and ordinary least squares regression do not perform well (i.e., do not provide accurate estimates) when data are highly skewed and contain outliers. Thus, the purpose of this dissertation was to investigate whether an alternative method, specifically the quantile regression model (QRM), outperforms traditional approaches during criterion-related validation and selection battery design. Across three unique samples, results suggest that although the QRM provides a much more detailed understanding of predictor-criterion relationships, the practical usefulness of the QRM in selection assessment battery design is similar to the OLS regression.
Recommended Citation
Theys, Evan, "" (2022). Dissertation. 962.
https://digitalcommons.latech.edu/dissertations/962