Looking for a relationship? Try GWR

It is often desirable in an analysis to examine the relationship between two or more variables. By relationship, we mean the manner in which one variable changes given change in another variable, ceteris paribus. An increase in the value of one variable might be associated with an increase in another; conversely, an increase in one variable might be associated with a decrease in another. It is very tempting when faced with a large number of variables, perhaps for thousands of observations, to reach for the principal components option in a statistical package and let the computer report the relationships it has found. Such a course of action would be abrogating one’s responsibility as an analyst to a computer and, more importantly, the resulting analysis would be partial, at best.