Point Features Rendering Scheme

June 04 2010 | 1 comment
Categories: Cartographic Design

I would like to get a clearer understanding of how best to use the Graduated Colours classification options for point features.

I have read the ArcGIS Help documentation and some other literature on it, and believe I understand the basics of Equal Interval, Quantile and Natural Breaks (though possibly not as clearly how the standard deviation classification works or how to interpret it since it doesn't seem to create divisions from mean, 1st, 2nd, 3rd standard deviation, but rather 0.5, 1.5, 2.5).

In what situations would you use one over the other for instance? E.g., Elevation point features

There's not a simple answer to this question.  Essentially this is what people start learning when they take a university level course in thematic cartography.  Unless you're already very familiar with how basic statistics relate to your reporting units, just buying a text book is likely to produce more questions than it answers.  However, the premise of your question is also flawed.  Typically one would not use graduated colors on points, as graduating the symbol's size is best way to denote quantity or magnitude at a given point location.

Point Features Rendering Scheme posted by Mathieu Cain on Jun 13 2010 10:16PM
I am well aware that:

For Equal Interval, each class has the same range of data covered (ranges being based on the entire dataset's range divided by the number of classes), which subsequently can lead to unequal number of values within each class.

For Quantile, each class has an equal number of values (intervals being based on the number of values divided by the number of classes), which can result in classes with unequal class ranges (so likely best suited for evenly distributed datasets, and at minimizing the effect of outliers).

For Natural Breaks, each class interval is determined using a statistical formula called Jenk's optimization that accounts for more inherent classifications in the dataset, though this could lead to unequal class ranges and number of values within each class (arguably, it is up to the user to visually determine the correct number of "natural" classes that occur).

For Standard Deviation, classes theoretically show the variation with respect to the mean of the dataset. I suppose this might only be used if the dataset approximated a normal distribution (resembling a bell shape, mean and median being practically the same, little to no skewness, etc.)

I still fail to understand why the standard deviation classification does not start at the mean (as 0) and create intervals at every standard deviation, as opposed to the half deviations that are used in the default.

I agree that there are various ways to represent a dot map, and that proportional symbol size is definitely another way to represent point features, however, while I may be completely off base, my inclination is that elevation data is much better suited represented as a graduated colour (much like a raster grid in a sense). That said, I am not convinced that there is one single graduated colour classification that is best suited for elevation data as this would likely largely depend on each individual dataset and how it was collected.