Cheng, Yongtian; (2025) Monte Carlo with Supervised Machine Learning Methods in Psychology: Simulations and Practice. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text ThesisYongtianCHENG.pdf - Accepted Version Download (703kB) | Preview

Abstract

Recent advancements in psychological research have increasingly integrated supervised machine learning methods to improve predictive accuracy. These models are generally developed using training/testing splits and cross-validation techniques. Some psychologists argue that this design encourages self-replication, suggesting that predictive outcomes derived from such models exhibit high reproducibility. However, empirical evidence supporting this claim within psychometric data is limited. This study addresses this gap by utilizing Monte Carlo simulations to assess the predictive performance of supervised neural networks across various psychometric datasets. A novel Monte Carlo simulation design was developed specifically for this investigation. The findings reveal that neural networks can enhance performance in binary prediction tasks involving random variables. However, the performance of supervised neural networks remains inconsistent, even with sample sizes as large as 10,000, and these models can generate seemingly meaningful results from random datasets. Based on the observed instability of neural network models, this thesis further explores an applied study using the elastic net method to predict aggressive behavior. The discussion addresses the implications of these findings, the limitations of the study, and recommendations for future research.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item