Bayesian Variable Selection Using EEG Data to Predict Transfer Performance After a Surprise Task Change

The purpose of this lab study was to demonstrate proof of concept for a novel Bayesian variable selection technique for predicting complex task performance. Specifically, we show results for a Bayesian variable selection technique that used the data (rather than a priori assumptions) to guide the selection of electroencephalographic (EEG) electrodes for predicting drops in performance scores after a surprise change in task demands. The sample consisted of eight young adults who completed 10 hr of training on a complex videogame across 4 days while wearing a 128-channel EEG net. Midway through Day 4, participants faced an unexpected increase in task complexity. The results indicated substantial variability across frequency ranges and time intervals in the electrodes predictive of performance. The results are discussed in terms of future research needs toward advancing statistical techniques to inform training practices for supporting trainees’ self-regulation in simulation-based training.