Analysis of GIS Data by Simple Examination - Introducing Geographic Information Systems with ArcGIS - page 379

A More Careful Look at Equal Intervals
13. Classify the values again. Choose the Equal Interval method again. Notice that the break lines
move back to the positions they had before. These positions are calculated so that each class
occupies the same distance along the horizontal axis. What is that distance? ________. The
calculation takes the form of “greatest value in the data set minus the least value, divided by
the number of classes.” Click OK, Apply, OK. Notice that the resulting map has five categories
in the T/C, but only four colors show up in the map, because one class is empty. What is the
range of the class that has no values in it? __________ - __________ Notice that the values
around 1000 and those around 5000 are lumped together. Look at the Classification window
again to notice that one class contains two clusters of values that beg to be separated, while
another class is empty. This is a difficulty with equal intervals.
Defined Interval
14. Choose the method Defined Interval. This is an equal interval method, but instead of the
interval size being chosen by the formula (High - Low)/N, it is chosen by the user and adjusted
so that the numbers look nice. Type 4000 in the Interval Size box and click the histogram. Note
that defining the interval determines the number of classes. How many? __________. Click OK.
Uncheck Show class ranges using feature values. Click Apply, then OK. Note the results on the
map and in the T/C. Again there are empty classes.
Quantiles
The problem of empty classes may be solved by the approach of quantiles. Quantiles places approximately
the same number of values in each category.
15. Classify and select the Quantile method. Change the number of classes back to 5 (this
gives you quintiles; other popular classifications are quartiles, deciles, and percentiles).
Click OK, Apply, and OK. Notice the map and the Table of Contents. Again, the pattern looks
somewhat similar to previous ones, but here the values around 1000 are broken into two
groups. A further complication: two of the values around 5000 are lumped in with some
1000s, while the other two are in with the 10000s and the 15000s. Again, the divisions seem
inappropriate.
Standard Deviation
The standard deviation method is not really appropriate for this set of data, because the values are not
distributed “normally”—that is, in a bell-shaped curve. But just because a method is inappropriate doesn't
mean we can't have the software apply it. It just means that we get bad results. (Garbage in, garbage out,
as the saying goes.)
16. Select the Standard Deviation method, and make the Interval Size one half of a standard
deviation. Find and click boxes to show both the Standard Deviation and the Mean on the
histogram (with vertical dashed lines). Click OK. Notice that the color ramp changes so that
the categories reflect a change at the mean (-0.25 to +0.25), rather than gradual change