Experiment: An Exercise Designed To Determine The Effects of One or
Experiment: An Exercise Designed To Determine The Effects of One or
1
1. 4. Experimental design
1. 4. 1. The role of experimental design
Experimental design concerns the validity and efficiency of the experiment.
The experimental design in the following diagram (Box et al., 1978), is
represented by a movable window through which certain aspects of the true
state of nature, more or less distorted by noise, may be observed. The
position and size of the window depend on the questions being asked by and
the quality of the experiment.
2
1.4.2. The selection of the design depends on the objectives and scope
The selection of an experimental design depends on your objective.
A null hypothesis (H0) can never be proven correct. It can only be rejected
with known risks of being wrong.
3
1. 4. 3. 1. Significance level of testing a hypothesis
Null hypothesis
Accepted Rejected
True Correct decision Type I error
Null hypothesis
False Type II error Correct decision
/2
Ho is true
Ho is false
1-
Acceptance Rejection
The only way to reduce both errors is to increase the number of replications
or improve the experimental design to reduce the error.
4
1. 4. 4. Specific issues of experimental design
Once the objectives, interesting questions, and the hypothesis are defined, the
scope, type, and requirements of an experiment are also more or less
determined. Thus, the experiment should be designed to meet those
requirements. Specifically, experimental design is concerned with the
following issues:
5
1. 4. 4. 1. Replication
Replication refers to the number of experimental units that are treated alike
(1) Each replication must be independent of every other.
(2) Each replication must be part of a randomized trial
1 2 3 4
5 6 7 8
9 10 11 12
6
Case 1: 4 plots are selected at random for each fertilizer.
B A C B
C A B A
C B A C
This is a completely randomized design. There are 4 replications for each
treatment. If the experiment is run in succeeding years then each year provides a
different set of replications.
Case 2: Same as case 1 except the crop is a perennial. The total yield of each
plot is measured at the end of each season.
In this case the succeeding years are not replications (not independent). This
is a repeated measures experiment.
Case 3: Same as case 1 except each of the 12 plots are further divided into
three subplots and yield is measured separately for each subplot.
B B B A A A C C C B B B
C C C A A A B B B A A A
C C C B B B A A A C C C
These subplots are not replications, they are subsamples. The randomization
requirement above is violated.
Case 4: The three treatment levels are randomly assigned to the three rows in
the field. Yield is measured on each plot.
B B B B
A A A A
C C C C
In this case each experimental unit is the row. The separate plots are
subsamples. This experiment has no replication.
7
1.4.4.2. Type and number of measurements
The theoretical consideration:
Y
n
The standard error determines the lengths of confidence intervals and the powers
of tests. Decreasing the standard error increases the precision of the experiment.
Precision has to do with the concept of random errors, and the precision of an
average can always be improved by increasing the sample size (n).
Finite resources.
Subsampling:
Multiple measurements on the same experimental unit yield a more precise value
for that experimental unit.
Subsampling can be used to reduce the apparent variation among e. u. subjected to
the same treatment and thereby reduce the standard error.
Y
n
8
The functions of randomization:
1. To neutralize systematic biases.
Proper randomization helps provide valid estimates of experimental
error and treatment means.
2. To ensure independence of errors, an assumption of many statistical
tests (ANOVA).
X X
X X X X
X X
X X X
X X X
Not accurate, not precise
Accurate, but not precise
XX
XX
X X XX
XX
X X X
X
9
1. 4. 4. 6. Relative precision of designs involving few treatments
I 1 / Y2 n / 2
If sY2 is used as an estimate of Y2 -> there is a correction factor:
I (n 1) /(n 3) sY2
Where s12 and s22 are the mean square errors (MSE) (from the ANOVA
Tables) of the first and second designs, and n1 and n2 are their degrees of
freedom of the two designs.
10