Agriculture Reference
In-Depth Information
The Kruskal-Wallis test is a nonparametric test with independent
samples where more than two medians are involved. It is similar to
the one-way ANOVA (analysis of variance) and has been referred to
as an analysis of variance with ranks.
The dataset Plant Flies.dta is a dataset of the number of flies per
square meter of foliage collected from a forest at different heights
(herbs, shrubs, and trees) (Zar, 1974, p. 140). Open this dataset and
enter the following:
kwallis
flies,
by(
plant
)
Kruskal-Wallis equality-of-populations rank test
+------------------------+
| plant | Obs | Rank Sum |
|-------+-----+----------|
| 1 | 4 | 41.00 |
| 2 | 4 | 23.00 |
| 3 | 4 | 14.00 |
+------------------------+
chi-squared = 7.269 with 2 d.f.
probability = 0.0264
chi-squared with ties = 7.269 with 2 d.f.
probability = 0.0264
The probability of 0.0264 indicates that there are differences in the
number of flies between the different strata. The medians are 10.85,
6.95, and 5.55 for the herbs, shrubs, and trees, respectively.
Let's examine another example using the Kruskal-Wallis test.
Open the dataset Rice Insecticides.dta, which is a CRD examining
different insecticide treatments to control brown planthoppers and
stem borers in rice (Gomez and Gomez, 1984, p. 14). This experiment
would normally be analyzed with a one-way ANOVA. In this case,
we are going to change a couple of the entries so that it includes some
ties. Change the Azodrin treatment with 2387 kg/ha to 2385 kg/ha
and change the Dol-Mix (1 kg) treatment with 2537 kg/ha to 2536
kg/ha. This gives us two values that are tied in the dataset. Enter the
following command and see the results:
Search WWH ::
Custom Search