Estimating the Unobserved: Hierarchical Bayesian Estimation of Evidence Accumulation Models with Missing or Contaminant Data

Abstract

This student project involved creating likelihood functions for the Bayesian estimation of cognitive parameters behind choices and response times in R, for the R package EMC2 of the Amsterdam Mathematical Psychology Lab. Particularly, the likelihood functions allow for good parameter recovery when data is missing (censoring and truncation), and when there are contaminant responses in the data. The improvement of censoring (taking the number of missing response times or choices into account) over truncation (discarding missing response times and choices) is tested in a fully reproducible simulation study.