Influence of Parsimony and Work-related Psychological Constructs in Predicting Turnover Intention when Using Machine Learning VS Regression - A Dissertation

Aug 5, 2026

This dissertation investigates the debate between traditional regression models and machine learning (ML) algorithms for predictive modeling, emphasizing how sample size and the number of variables affect accuracy. Study 1 examines whether ML offers advantages over regression as sample size grows and how variable dimensionality shapes the trade-off between approaches. Study 2 evaluates incremental validity by testing whether work-related psychological constructs improve ML prediction of turnover intention beyond biodata alone. Analyses use Federal Employee Viewpoint Survey data and an online MTurk dataset, comparing Gradient Boosting Trees, Random Forest, Neural Networks, and Support Vector Machines with linear and logistic regression.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages154

Reading Time6 h 28 min

AccessFree to download

Introduction

What Constitutes Machine Learning

Theory-based models

Empirical-based models

Common Terms in Machine Learning

Data Hungriness

Parsimony

Algorithms

Decision Trees

Classification and Regression Trees (CART)

Random Forests (RF)

Gradient Boosting Trees (GBT)

Bayesian Additive Regression Trees (BART)

Neural Networks (NN)

Support Vector Machines (SVM) classifiers

5-2024 Influence of Parsimony and Work-related Psychological Constructs in Predicting Turnover Intention when Using MachineLearning VS Regression Diego Figueiras Influence of Parsimony and Work-related Psychological Constructs in Predicting TurnoverIntention when Using Machine Learning VS Regression A DISSERTATION Submitted to the Faculty ofMontclair State University in partial fulfillmentof the requirementsfor the degree of Doctor of Philosophy byDiego FigueirasMontclair State UniversityMontclair,NJMay 2024 INFLUENCE OF PARSIMONY Graduate Program:Industrial/Organizational Psychology Certified by: Dr.Kenneth Sumner Associate Provost for Academic Affairsand Acting Dean of the Graduate School 5/1/24 MONTCLAIR STATE UNIVERSITYTHE GRADUATE SCHOOLDISSERTATION APPROVAL We hereby approve the Dissertation Influence of Parsimony and Work-related Psychological Constructs in PredictingTurnover Intention when Using Machine Learning VS Regression of Diego FigueirasCandidate for the Degree:Doctor of Philosophy Dissertation Committee: Dr.Michael BixterDissertation Chair Dr.Kevin Askew Dr.John T.Kulas INFLUENCE OF PARSIMONY Copyright@2024by Diego Figueiras.All rights reserved. INFLUENCE OF PARSIMONY # Abstract This dissertation explores the ongoing debate between traditional statistical regressionmodels and machine learning(ML)algorithms in predictive modeling,focusing on the impact ofsample size and the number of variables.Study 1 investigates the relationship between sample sizeand predictive accuracy,proposing hypotheses regarding the advantages of MLover regression assample size increases.Additionally,the study examines the influence of the number of variables onpredictive accuracy,emphasizing the trade-off between ML and regression models.Using datafrom the Federal Employee Viewpoint Survey,the research aims to contribute insights into theconditions favoring each modeling approach.Study 2 shifts the focus to incremental validity,exploring whether work-related psychological constructs enhance ML models'predictive accuracyin turnover intention compared to biodata alone.The proposed hypotheses suggest that incorporating psychological constructs will improve predictive accuracy,addressing the “garbage ingarbage out"concern prevalent in ML applications.The methods involve diverse datasets,including responses from federal employees and an online survey through Amazon's MTurk,withmachine learning algorithms such as Gradient Boosting Trees,Random Forest,Neural Networks,and Support Vector Machines being compared to linear and logistic regressions.The dissertationseeks to advance understanding in the field,offering practical insights for researchers andpractitioners navigating the dynamic landscape of predictive modeling. Keywords:turnover intention,machine learning,regression,parsimony,biodata # Acknowledgments The following dissertation study would not have been possible without the financial,administrative,technical,and social support of the Psychology Department and Graduate School atMontclair State University(MSU).It has been a great experience going from the more impersonalenvironment of my undergraduate university to being part of a tight team of smart people. I want to thank Dr.Bixter,the chair of my committee.I first met him when taking a coursein multivariate statistics.I was a bit scared of the name alone,but Dr.Bixter made it fun.We hadtons of hands down practice using SPSS and R,something that many statistics and even computerscience instructors forget to do.He has taught me more than I could ever give him credit for here.Next,I'd like to thank Dr.Kulas,my psychometrics professor and career mentor over the past fiveyears.He showed me that R,the annoying,"worse than python"programming language I onceused in a statistics class during my undergraduate was actually pretty useful.This entire manuscriptwas built in RStudio,and it would not have been possible without his constant

FAQs

How does the dissertation compare machine learning and regression models for predicting turnover intention?

It contrasts ML algorithms with linear and logistic regressions, evaluating predictive accuracy under different data conditions and modeling inputs.

What factors are examined in Study 1 regarding ML versus regression performance?

Study 1 focuses on the effects of sample size and the number of variables on predictive accuracy, proposing hypotheses about where ML may outperform regression.

How does Study 2 assess the value of work-related psychological constructs?

Study 2 tests incremental validity by determining whether psychological constructs enhance ML prediction of turnover intention beyond biodata alone, addressing concerns about “garbage in garbage out.”

Aug 5, 2026

Upload by Theodore