CFE 2015: Start Registration
View Submission - CFE
A1056
Title: Estimation of continuous piecewise linear models using Bayesian clustering Authors:  Fabrizio Carmignani - Griffith University (Australia)
Eric Eisenstat - The University of Queensland (Australia) [presenting]
Rodney Strachan - The University of Queensland (Australia)
Rabee Tourky - Australian National University (Australia)
Abstract: We develop methods for efficiently estimating continuous piecewise linear functions based on a Bayesian approach to clustering. Specifically, we design MCMC algorithms that rely on clustering to generate hyperplanes and combine these with the Reisz estimator to construct a joint M-H proposal. The probabilistic approach circumvents the need to search exhaustively through a vast number of possible hyperplanes and thus surmounts the curse of dimensionality associated with existing methods. We use our method to estimate a nonlinear growth equation, nested within a VAR involving fiscal policy, monetary policy and output growth.