Stochastic Conditional Duration Models with Mixture Processes

This paper studies stochastic conditional duration models with a mixture of distribution processes for financial asset's transaction data. The mixture component distributions include exponential, gamma and Weibull. The models allow for a correlation between the observed durations and the logarithm of the conditional expected durations. Suitable MCMC algorithms are developed for Bayesian inference of parameters and duration forecasting of the models. Unlike much of the existing studies in this literature, simulation studies and empirical applications suggest that the proposed models and method are able to t the left tail of the marginal distribution of duration time series relatively well