ABSTRACT
The chapter on univariate and bivariate data exploration reviews classic graphs used in exploratory data analysis (EDA) and their implementation in GeoDa. Particular attention is paid to spatializing such techniques by exploiting linking with a map and dynamic brushing.
Basic statistical graphs include the histogram, box plot and scatter plot. For the latter, different smoothing options are reviewed, such as a linear fit and a LOWESS (local regression) fit. Scatter plot brushing is illustrated, with a particular focus on detecting spatial heterogeneity by means of a Chow test that is dynamically recalculated in GeoDa for each selection in the scatter plot.
A series of bivariate associations is visualized in the scatter plot matrix, which again allows for both linear and LOWESS smoothing of the scatter plot cells.
In addition to investigating spatial heterogeneity by means of brushing a linked scatter plot and map, GeoDa's averages chart provides a rudimentary way to analyze and visualize treatment effects. In essence, it consists of a difference in means test, where the selection of treated and controls is implemented through selection on a map.
