Modern Prediction Methods: New Perspectives on a Common Problem. (July 2018)

• Record Type: Journal Article
• Authors: Putka, Dan J.; Beatty, Adam S.; Reeder, Matthew C.
• Other Names: LeBreton James M. guest-editor.; Meade Adam W. guest-editor.
• Is Part Of: Organizational research methods. Volume 21:Number 3(2018:Jul.)
• Page Start: 689
• Page End: 732
• Publication Date: 2018-07
• Subjects: prediction -- lasso -- least angle regression -- elastic nets -- classification and regression trees -- bagged trees -- random forests -- gradient boosted trees -- support vector machines -- bagging -- boosting -- bias-variance tradeoff -- cross-validation -- model tuning -- biodata; Organization -- Research -- Methodology -- Periodicals; Organizational sociology -- Research -- Methodology -- Periodicals; Management -- Research -- Methodology -- Periodicals; 302.350721
• DOI: 10.1177/1094428117697041
• Languages: English
• ISSNs: 1094-4281

Abstract

Predicting outcomes is critical in many domains of organizational research and practice. Over the past few decades, there have been substantial advances in predictive modeling methods and concepts from the computer science, machine learning, and statistics literatures that may have potential value for organizational science and practice. Nevertheless, treatment of these modern methods in major management and industrial-organizational psychology journals remains minimal. The purpose of this article is to (a) raise awareness among organizational researchers and practitioners with regard to several modern prediction methods and concepts, (b) discuss in nonmathematical terms how they compare to traditional regression-based prediction methods, and (c) provide an empirical example of their application and performance relative to traditional methods. Beyond illustrating their potential for improving prediction, we will also illustrate how these methods can offer deeper insights into how predictor content functions beyond simple construct-based explanations.