Agriculture Reference
In-Depth Information
in each rep, and the Residual is the error or background noise that
occurs in the experiment. The Residual is important in this design
because it is the denominator in the calculated F-tests. The only really
important F-test is for entry where we see a highly significant differ-
ence between varieties with a Prob > F of 0.0000.
Because the rep source of variation was not significant, there is not
much difference between calculating this model as an RCBD or a
CRD (completely randomized design). This is not always the case; the
blocking effect (rep), when significant, can account for a lot of varia-
tion in the model. This accounted-for variation can lower the Residual
mean square making it more likely to detect differences between the
treatments. In fact, it is possible to calculate the relative efficiency of
RCBD compared to a CRD by the formula
(
)
(
)
r
1
Ert
+−
1
E
b
e
RE
..=
(
)
rt
1
E
e
In this formula, the r represents the number of replications, which is
3 in this case. E b is the replication mean square, which is 10990.0706,
and E e is the residual mean square, which is 10294.9215. The t is the
number of treatments, which in this case is 20.
In Chapter 3, it was mentioned that some commands save results
for further calculations. The summarize command was used as an
example saving several results in r() . he anova command also saves
results, but these results are saved in e() , which is used by e-class
commands, estimation commands. Type in ereturn list , which
should be entered immediately after the anova command:
ereturn list
The following results will appear:
scalars:
e(N) = 60
e(df_m) = 21
e(df_r) = 38
e(F) = 6.734516384698626
e(r2) = .7882121609461262
e(rmse) = 101.4638926803575
e(mss) = 1455957.670464247
Search WWH ::




Custom Search