Two-Group versus Multiple-Group Analysis
In a simple statistical analysis we may be characterizing one
group of data or comparing that group with a standard value. It is slightly more
intricate to compare two groups when we want to characterize each group and determine
whether the groups are statistically different. Much more concern must be applied
to the situation when we have three or more groups. In this case of multiple groups,
two distinct types of questions can be asked. First, we might wish to know whether
the groups are significantly different or whether they are all the same (with any
numerical differences attributed solely to random fluctuations). Second, if we do
conclude that the groups are statistically not all the same, we might want to know
which groups are different from one another—that is, are they all distinct
or are just one or a few different whereas the rest are similar to each other.
In summary, three questions should be asked when choosing between
tests (see Table 23-3
):
- One group, two groups, or many groups?
- Paired or unpaired groups?
- Parametric (interval data following a normal distribution) or nonparametric
methods?