Agriculture Reference
In-Depth Information
8
P oSt H oc t eStS
Planned Comparisons
Analysis of variance will answer the question: Are there significant
differences between treatments and are there any interactions between
factors when more than one factor is involved? Post hoc tests are per-
formed after the ANOVA (analysis of variance) to answer the specific
question about which treatments differ.
Oftentimes the experiment is such that logical comparisons between
the treatments can be planned and evaluated. It is generally not rec-
ommended that comparisons be chosen based on the results. The
comparisons of interest should be considered and planned in advance
to avoid any bias. Load the dataset Rice Fertilizer Comparisons.dta
and enter the command
anova yield trt rep
This dataset is of an experiment on rice yield with different fertil-
izers including ammonium sulfate (NH 4 SO 4 ), green leaf (presumably
some type of compost or organic matter), and a combination of both
(Palaniswamy and Palaniswamy, 2006, p. 401). The ANOVA results
are significant and there are comparisons of interest that were decided
on in advance of the experiment. These include comparing the control
to the fertilizer treatments, comparing NH 4 SO 4 to the green leaf, and
comparing the NH 4 SO 4 and green leaf treatments to the combination
of NH 4 SO 4 and green leaf. TableĀ 8.1 presents these results as orthog-
onal coefficients. Comparisons are considered orthogonal when the
coefficients add to 0.
Immediately after the ANOVA, the test command can be used to
make these specific single degree of freedom comparisons. The com-
mands and output for the first comparison, control versus fertilization,
are
147
Search WWH ::




Custom Search