Agriculture Reference
In-Depth Information
(continued)
(C) ANOVA with change of origin
Variety 1
Variety 2
Variety 3
Variety 4
ANOVA
SOV
d.f.
SS
MS
F
P
value
Irrigation
2
143.704
71.852
486.903
0.000
Variety
3
668.164
222.721
1509.266
0.000
Interaction (I
V)
6
10.761
1.794
12.154
0.000
Error
24
3.542
0.148
Total
35
826.171
From the above analyses the following points
may be noted:
1.
and fast rule to determine the cutoff value of CV
% for an experiment is reliable; it depends on the
condition of the experiment (laboratory condi-
tion, field condition, etc.), type of materials/
treatments tested in the experiment, desired pre-
cision from the experiment, etc. Generally the
CV% should be less in experiments conducted
under laboratory conditions compared to field
experiments. Similarly CV% also depends on
the type of field crop, size and shape of experi-
mental units, etc. However, by and large, a CV%
less than 20% is regarded as an indication for the
reliability of the experiments. If the CV value is
more than 20%, there is a need to verify the
experimental procedure and need for emphasis
on the reduction of experimental error.
-ratios and the significance level do not
change under the above three cases.
2. Mean values change in the second and third
cases.
3. Sumof squares andmean sumof squares values
do not change in the first and second cases.
4. As error mean square remains the same in the
first and second cases, critical difference
values also remain the same.
Thus, with the change of origin and/or scale,
the basic conclusion from ANOVA does not
change, but while comparing the means, care
should be taken to adjust the mean values and
the critical difference values accordingly.
F
10.5
Experimental Reliability
10.6
Comparison of Means
The
-test in analysis of variance indicates the
acceptance or the rejection of the null hypothesis.
The
F
Reliability of experimental procedure, in its
simplest form, can be verified with the help of
coefficient of variations. The coefficient of vari-
ation is defined as
F-test in analysis of variance can be signi-
ficant even if only one pair of means among
several pairs of means is significant. If the null
hypothesis is rejected, we need to find out the
means which are significantly different from
each other and resulting the significance of
F-test
D
Grand Mean
S
:
CV
¼
100
:
Here the positive square root of error mean
square in analysis of variance is taken as an
estimate of the standard deviation. Thus, from
the table of analysis of variance, one can work
out the CV% as follows:
. Thus, comparison of treatment means
becomes essential. The comparison of treatment
means can be done either through
pair compari-
son
. In group com-
parison, two or more than two treatment means
are involved in the process. Under the group
comparison there are
or through
group comparison
p
Grand Mean
ErMS
CV
¼
100
:
between-group compari-
son, within-group comparison, trend compari-
son, and factorial comparison.
Now the question is, what should be the range
or the value of CV% for the information from an
experiment to be taken reliable? There is no hard
Pair comparison
is the simplest and most commonly used method.
