Tree-Structured Histogram Estimation Based on a Simple Class of Wavelets
Joachim Engel · RePEc: Research Papers in Economics · 1993
Adaptive multivariate histograms that are obtained through recursive partitioning of the sample space are studied. The central idea is to consider the tree-structured histogram as a wavelet estimator based on the orthogonal system of Haar functions. From the formula for the mean squared error a spatially adaptive data-driven smoothing method is proposed leading to an unequal bin width histogram.