© Copyright StatSoft, Inc., 1984-2003
Statistical Advisor
More Info
The distinction between options 1/2 and 3/4 (general nonparametric comparison between
groups) should be based on the following considerations:
- Is the variable (or variables) on which the comparison is to be based measured on at
least an interval scale level? Interval scale measurement means that, for example, a
difference between the values 1 and 2 in that variable is comparable to the difference
between 3 and 4. Often, the values of variables contain only rank order information, and
do not allow us to compare intervals. Refer to the manual (Elementary Concepts) for
more details concerning this issue.
- Does the distribution of data points follow (approximately) the normal distribution?
You can easily check this condition by viewing a histogram.
- Is the data set reasonably large? For example, are there at least 20 observations in
the comparison?
If any of these three conditions are not met, it is advisable to verify any results via
the appropriate nonparametric tests discussed in NONPARAMETRICS &
DISTRIBUTIONS.
© Copyright StatSoft, Inc., 1984-2003
STATISTICA is a trademark of StatSoft, Inc.