摘要：Imposing curvature and monotonocity restrictions always has been a vexing problem for applied economists. We present an estimation methodology for more efficiently estimating flexible functional forms subject to "regional", as opposed to "local" or "global", regularity conditions. The technique involves a Markov Chain Monte Carlo sampler to ensure consistency over a predefined connected set of the regressor space. Numerical examples demonstrate the advantages and technical feasibility of our new approach.
