Joint Quantile Regression through Bayesian Semiparametrics

Surya T. Tokdar and Joseph B. Kadane


We introduce a Bayesian semiparametric methodology for joint quantile regression with linearity and piecewise linearity constraints. We develop a probability model for all quantile curves in a continuum that define a coherent sampling distribution of the response variable. We provide a detailed illustration of model fitting and inference by analyzing wind speed trends of tropical cyclones in the North Atlantic.

Keywords: Bayesian inference, Gaussian process, Tropical cyclones, Change points, Markov chain Monte Carlo.

Heidi Sestrich 2009-07-27
Here is the full PDF text for this technical report. It is 1801479 bytes long.