Bayesian inference on time-varying proportions

摘要: Time-varying proportions arise frequently in economics. Market shares show the relative importance of firms a market. Labor economists divide populations into different labor market segments. Expenditure describe how consumers and allocate total expenditure to various categories. We introduce state space model where unobserved states are Gaussian observations conditionally Dirichlet. Markov chain Monte Carlo techniques allow inference for unknown parameters states. draw as block using multivariate proposal distribution based on quadratic approximation log conditional density given data. Repeated draws from particularly efficient. illustrate automobile production