Agriculture Reference
In-Depth Information
There are a number of examples of using a covariate to eliminate
some factor that is affecting the outcome and to more precisely cal-
culate the residual. One is stand count, which is used as a common
covariate in many experiments. This is useful where the plant stand
is not complete because of poor germination or adverse effects after
transplanting.
Other examples of covariate use involve the initial weight of exper-
imental animals where the gain in weight is the dependent variable
and this gain in weight may be affected by the animal's initial weight.
Field position also may be used as a covariate. For example, an
experiment may have been planted near the edge of a field where there
is a distinct edge effect. Perhaps plants along the field's edge may be
robbing nutrients and water from your experimental plants, which, in
turn, could affect your results. In this case, using the reciprocal of the
distance from the field's edge would be an appropriate covariate. By
using such a covariate, the farther from the field's edge, the lower the
effect. This type of effect would generally be taken care of by block-
ing, but in some cases such effects may not be completely evident at
the start of an experiment.
Another example is soil heterogeneity and its effect on treatment
effects. Blocking, as in an RCBD, can have a significant impact in
reducing effects due to plot location in a field. But sometimes soils
can be heterogeneous in such a way that blocking cannot easily deal
with the problem. A uniformity trial prior to experimental work
can identify such soil heterogeneity and these data can be used as a
covariate.
Analysis of covariance can be used to estimate missing data as well.
To estimate missing data and complete the analysis, first set the miss-
ing data point to 0, then set up a covariate that has values of 0 for all
data points except the one with the missing value, which should be set
to 1. Then conduct the analysis of covariance. To see this, we will use
the Covmissing.dta dataset. This is a dataset of ascorbic acid content
in turnip greens with three treatments of postharvest handling (Steel
and Torrie, 1980, p. 427). Replace the missing data point with a 0.
Then create a new covariate (i.e., X) with values of 0 for all entries
except for the data point that is missing, which will have a value of
1. This is often referred to as a dummy variable. The following com-
mands will accomplish this:
Search WWH ::
Custom Search