A general latent assignment approach for modeling psychological contaminants

作者：Zeigenfuse Matthew D^*; Lee Michael D

来源：Journal of Mathematical Psychology, 2010, 54(4): 352-362.

DOI：10.1016/j.jmp.2010.04.001

Data from psychological experiments are rife with 'contaminants', which can generally be defined as data generated by psychological processes different from those intended as the object of study. Contaminant data can interfere with the testing of substantive psychological models and their parameters, so It is important to have methods for their identification and removal. After noting that current practices in cognitive modeling for dealing with contaminants are not completely satisfactory, we argue for a general latent mixture approach to the problem We demonstrate the tractability and effectiveness of the approach concretely, through a series of four applications. These applications involve a simple choice problem, a diffusion model of a response time and accuracy in decision-making, a hierarchical signal detection model of recognition memory, and a reinforcement learning model of decision-making on bandit problems. We conclude that developing models of contaminant processes requires the same sort of creative effort that is needed to model subs

• 出版日期2010-8