The Accuracy of Spatial Disaggregation Technique – Entropy Based Distributionally Weighted Regression, Simulation Study of Several Types of Predictor and Weight Variable
The disaggregation technique must be applied in certain cases when information only available at a high level spatial unit, to draw any conclusion for the lower level spatial unit. This problem commonly occurs in the area of regional economics, agricultural economics or environment study. One of the disaggregation techniques, namely entropy based distributionally weighted regression is particularly the main interest in this study. It conducts a simulation study to analyze the effect of the degree of the explanatory power of the predictor, and the type of correlation between response and weight variable, on the accuracy of the method. The simulation study shows that the technique works well for predictor with at least medium explanatory power. The use of negatively correlated weight will not lead to a serious accuracy problem, as long as the correlation is kept at a low degree.
Keywords - Spatial disaggregation, entropy maximization, distributionally weighted regression, ecological inference, MAUP