Page 272 - Area 9 - Relevant Document

19

does not use randomization then he cannot tell whether observed
differences are due to differences in treatment, or due to the systematic
assignment of treatment. Another basic reason is to have valid measure
of experimental error. If a systematic bias is present in the experiment,
then estimated error variances are not only due to the inherent
variability of the experimental units but also due to the systematic bias,
which is combined into. This term will reduce the reliability of the test.

3. Control of error may be introduced in the different stages of the
experiment besides the
Inherent variability of the experimental units used. Hence, the
researcher should have proper knowledge of the experimental units he
is going to use whether these units’ needs to be groups called blocks
and most importantly, he has the control over all factors that may
contaminate the experiment.

Suggested ways of minimizing error in the experiment:

1. Increase the size of the experiment either through the provision of
more replicate or by inclusion of additional treatments.
2. Selection of treatments
3. Proper choice of experimental design
4. Refinement of experimental techniques

4. Number of replications and selection of treatments
The number of replication is important in estimating the standard error
of the mean or the difference of the two mean. If the experiment consists
only one replicate, no estimate of experimental error is available
therefore there is no way to determine whether observed differences
indicate real or differences due to inherent variation. According to the
Central Limit Theorem: as the number of replication increases, the
estimate of the population becomes more precise.

5. Selection of treatments
In some field experiments, the rate or the amount of fertilizer is dictated
by the objective of the research. If one of the research objectives is to
determine the optimum amount of fertilizer (in fertilizer trial) that would
give maximum response, then not only two or three rates of fertilizer
must be included but more than three, the reason is that if only two
rates of fertilizer are included, then the trend line that can be produced
will be only linear.