Model Comparison and Parameter Recovery of Two Diffusion Models: Fast-dm and HDDM
Yueyi Jiang · 2022
Diffusion models (Ratcliff, 1978) can be used to make inferences about the ongoing cognitive processes in fast binary decision task performance. In recent years, different techniques have been developed to fit the diffusion model to experimental data. The present paper compared the model performance of the two popular methods in estimating parameters of the diffusion model, fast-dm (fast diffusion model) and HDDM (hierarchical drift diffusion model). Our results indicated that fast-dm and HDDM models are systematically different in parameter estimation both at a group level and at an individual level across experimental conditions on the recognition memory task performance. We found that the observed differences of the two model fitting methods were not due to the inconsistency in parameter estimation. In general, fast-dm performed better in model fitting than HDDM. Moreover, our findings support the interpretations of the diffusion model parameters about the ongoing cognitive processes of the recognition memory task performance. These findings can provide a basis in understanding the strengths and limits of the application of diffusion model.