PEP 0003 Spatial Smoothing Module

Author	Myunghwa Hwang <mhwang4@gmail.com> Luc Anselin <luc.anselin@asu.edu> Serge Rey <srey@asu.edu>
Status	Draft
Created	11-Feb-2010
Updated

Abstract

Spatial smoothing techniques aim to adjust problems with applying simple normalization to rate computation. Geographic studies of disease widely adopt these techniques to better summarize spatial patterns of disease occurrences. The smoothing module combines a number of previously developed and to-be-developed classes for carrying out spatial smoothing. It will include classes for the following techniques: mean and median based smoothing, nonparametric smoothing, and empirical Bayes smoothing.

Motivation

Despite wide usage of spatial smoothing techniques in epidemiology, there are only few software libraries that include a range of different smoothing techniques at one place. Since spatial smoothing is a subtype of exploratory data analysis method, PySAL is the best place that host multiple smoothing techniques.

The smoothing module will mainly implement the techniques reported in [Anselin2006].

Reference Implementation

We suggest adding the module pysal.esda.smoothing which in turn would encompass the following modules:

• locally weighted averages, locally weighted median, headbanging
• spatial rate smoothing
• excess risk, empricial Bayes smoothing, spatial empirical Bayes smoothing
• headbanging

References

[Anselin2006] Anselin, L., N. Lozano, and J. Koschinsky (2006) Rate Transformations and Smoothing, GeoDa Center Research Report.