ANOVA (F-statistic)
and Kruskal-Wallis (H-statistic)
The one-way ANOVA has several assumptions:
- Dependent measure is interval; independent is nominal
- Normal distrubtion of data in dependent measure
- Homogeneity of varience in dependent measure
- Independence of groups
The F-statistic results from performing ANOVA.
Consider the evaluation data in the evals.sav file. For the analysis on whether students' evaluations of instructors
are determined by the students' expected grades, the dependent variable
is not interval. As a result, it is not legitimate to examine this
question with ANOVA.
An applicable statistic is the Kruskal-Wallis,
which has the following assumptions:
- Dependent measure is either ordinal or interval, but not completely normally
distributed
- All groups are independent
- Groups are similar in shape
Kruskal-Wallis reports an H-statistic (interpretable in much the same way
that the F-statistic is interpreted). Note that Kruskal-Wallis does not
require interval data. To determine the general shape of data, either examine them
graphically or perform descriptive
statistics on the two groups to make sure they are similar.
To access Kruskall-Wallis in SPSS,
- Select "Analyze" then "Nonparametric Tests" then
"K Independent Samples"
- Select one or more numeric variables
- Select a grouping variable
- Click "Define Range"
to specify minimum and maximum integer values for the grouping variable.