Choropleth visualisation to show variations from correlation I'm trying to think of a good way to visualise a dataset which will consist of:


*

*a region identifier (for which I have geographies)

*a socioeconomic index expressed as a decile (this region is in the 30% most disadvantaged). Call it S.

*several other metrics we compute, which boil down to how much funding this region is receiving. For the sake of this question, let's focus on the metric "dollars per capita", called F.


Now, from the perspective of the user, there is a preferred negative correlation between the socioeconomic index and the funding metric: more disadvantaged areas should receive more dollars per capita.
So: what is a good metric derived from S and F that would yield a choropleth in which departures from the preferred correlation are highlighted? That is, an advantaged region with low funding, or a disadvantaged region with high funding should get a sort of "neutral" score, while the other two cases should be opposite extremes.
I don't have a stats background, in case this isn't obvious. :) The final solution will have to be implemented in SQL and JavaScript.
I suspect principal component analysis is what I need, but a quick and dirty alternative would be great.
 A: First of all, a choropleth may not yield a truthful or even useful visualization for your task because region area size would not necessarily correlate well with funding size and the socioeconomic index. 
A better alternative would be a Dorling cartogram. These days, D3.js or Vega.js can be used for implementation in JavaScript - the above example was implemented using the older protovis.
For the sole purpose of visualisation, this also answers your original question because all of a sudden you have two valid encoding channels at your disposal instead of just one: size, for either funding budget or socioeconomic index, and color / brightness. Since both variables are quantitative, you will have to decide which variable to encode with color / brightness and whether to bin the said variable or to use a gradient scale - a bit of experimentation should shed some light on this minor issue. 
It appears from your post that you may have several variables for measuring funding size. Here, an analysis of correlation with the socioeconomic index would help to select the most appropriate one(s). The other road would be dimension reduction using PCA.
