Using HMC Approach To Evaluate Parameters Of IRT Models

How to obtain more accurate model parameter estimation results in the case of complex models is a classic research direction of psychological measurement.In this study,a relatively new Markov chain Monte Carlo method is introduced: Hamiltonian Monte Carlo(HMC)is introduced into the parameter estimation of the threeparameter multidimensional logistic model.For this algorithm,this study proposes to solve 1)as a Markov chain Monte Carlo method,how many iterations can the HMC converge when estimating the item response theory model;2)whether the HMC is better than previous methods;3)Whether the method can obtain robust parameter estimation results.This study proves through three simulation studies that 1)the Markov chain Monte Carlo method can reach the convergence criterion faster and gives the proposed chain length setting;2)By comparing several common parameter estimates method,the performance of the HMC method in the parameter estimation of the multidimensional project response theory model is acceptable.The upper asymptotic parameter estimation of the three-parameter logistic model is excellent;3)the method is less affected by the prior information and less than the current mainstream method.Combining all the characteristics,the HMC is a good item response theory model parameter estimation method.