Accepted Solutions
there are many classification schemes, for example std deviations.
I prefer to produce a classification scheme that is derived from the histogram of the actual data, since any false presumptions of normalcy in the data can be dealt with. If you are interested in between region variations then I would use one scheme. If you are interest at within region variations, then I see the point of a separate one for each region, but I would also advise you to provide the measure of centrality and dispersion the label or within the legend so that people are forewarned that the visual pattern for the with region data is based upon these measures and their absolute values should not be directly compared to other regions. Should wish to normalize the scheme based on their measures, I would still report them. People sometimes get complacent when making visual comparison without knowing what the underlying statistical measures are (ie two regions showing a non-clustered normal looking pattern, can differ markedly in their fundamental values) In short... lie with statistics and lie with maps should be avoided at all costs 
Indeed, I'm referring to the quantiles symbology. This is quite useful to understanding relationships between areas and their values.
Well, the comparison will be more within the state than across states, that's why I need the "quantilization" to be done by municipalities...
As all the values go from 0 to 1, I don't see another classiication scheme to use. Any ideas?
Here's an example I did by CLIP the municipalities to the state:
Nationwide: 
Only one state: 
Much clearer classification and what I'm looking for, but impractical for 150 states as this will leave me with 150 layers!
there are many classification schemes, for example std deviations.
I prefer to produce a classification scheme that is derived from the histogram of the actual data, since any false presumptions of normalcy in the data can be dealt with. If you are interested in between region variations then I would use one scheme. If you are interest at within region variations, then I see the point of a separate one for each region, but I would also advise you to provide the measure of centrality and dispersion the label or within the legend so that people are forewarned that the visual pattern for the with region data is based upon these measures and their absolute values should not be directly compared to other regions. Should wish to normalize the scheme based on their measures, I would still report them. People sometimes get complacent when making visual comparison without knowing what the underlying statistical measures are (ie two regions showing a non-clustered normal looking pattern, can differ markedly in their fundamental values) In short... lie with statistics and lie with maps should be avoided at all costs
