Modeling Spatial, Temporal, and Multivariate Autoregressive Dependence in Real Estate Prices
摘要
We use parallel processing and improved algorithms to search for the weights given to two locational coordinates as well as three non-locational dimensions to find multidimensional neighbors (previous in time) to fit a multidimensional STAR model. We find that the improvements allow for quick specification searches. The effect of the improved specifications is a material reduction in the standard deviation of the residuals.