User Guide

126
Chapter 6
Equal Percentiles Based on S canned Cases. Generates banded categories with an
equal number of cases in each band (using the aempirical algorithm for percentiles),
basedoneit
her of the following criteria:
 Number of C
utpoints.
The number of banded categories is the number of cutpoints
plus one. For example, three cutpoints generate four percentile bands (quartiles),
each containing 25% of the cases.
 Width (%). Width of each interval, expressed as a percentage of the total number
of cases. Fo
r example, a value of 33.3 would produce three banded categories
(two cutpoints), each containing 33.3% of the cases.
If the source variable contains a relatively small number of distinct values or a large
number of c
ases with the same value, you may get fewer bands than requested. If
there are multiple identical values at a cutpoint, they will all go into the same interval;
so the actual percentages may not always be exactly equal.
Cutpoints at Mean and Selected Standard Deviations Based on Scanned Cases.
Generate
s banded categories based on the values of the mean and standard deviation
of the distribution of the variable.
 If you don’t select any of the standard deviation intervals, two banded categories
will be cr
eated, with the mean as the cutpoint dividing the bands.
 You can s
elect any combination of standard deviation intervals based on one, two,
and/or three standard deviations. For example, selecting all three would result in
eight banded categories—six bands in one standard deviation intervals and two
bands fo
r cases more than three standard deviations above and below the mean.
In a norm
al distribution, 68% of the cases fall within one standard deviation of
the mean, 95%, within two standard deviations, and 99%, within three standard
deviations. Creating banded categories based on standard deviations may result in
some de
fined bands outside the actual data range and even outside the range of
possible data values (for example, a negative salary range).
Note: Calculations of percentiles and standard deviations are based on the scanned
cases.
If you limit the number of cases scanned, the resulting bands may not contain
the proportion of cases that you wanted in those bands, particularly if the data file is
sorted by the source variable. For example, if you limit the scan to the first 100 cases
of a da
ta file with 1000 cases and the data file is sorted in ascending order of age of
respondent, instead of four percentile age bands each containing 25% of the cases,