APH 517: Animal Production Research Techniques

EXPERIMENTAL DESIGNS, REPORTING AND THE USE OF COMPUTER

FOR DATA ANALYSES IN ANIMAL PRODUCTION RESEARCHES

Prof. Deji Fasuyi

EXPERIMENTAL DESIGNS, REPORTING AND THE USE OF COMPUTER
FOR DATA ANALYSES IN ANIMAL PRODUCTION RESEARCHES

Introduction
Designing and analyzing of experiments are very important tools in Animal Production experiments.
These are various steps involved in the use of statistical methods in the planning and execution of a
research study to ensure that important and valid experimental data are collected for analysis and to
generate adequate conclusions. These statistical methods have wide application in various fields of
human endeavours and generally known as biometrics in life sciences including agriculture. A sound
mastery of biometrics promotes a clear analytical thinking required for objective evaluation of research
findings and hypotheses derived therein.

Matters that naturally arise from biometrics in animal production research works include

1. The field layout and experimental design to be adopted

2. The type of data or parameters to be examined and the mode of collecting such
3. The suitable formatting method of collected data for clear recording, summarizing and
presentation
4. The computations and tests of significance to be analysed.

There is no doubt that a thorough knowledge of the subject matter in animal production experiment is
the panacea for a good research study and the statistical tools employed is a facility used to arrive at
adequately reliable and presentable results.

Experimental Designs
All preliminary steps taken to ensure that appropriate data are obtained from an experimental study to
facilitate correct analysis and to generate valid inferences are collectively referred to as experimental
design. Experimental designs starts when an experiment is formulated and ends when all data have
been collected for analysis. Experimental design refers to the patterns of arranging the experimental
materials which include the experimental units, sampling units and the treatments.

The experimental unit is the material to which a single treatment is applied in one replication of the
basic experiment. For example, an animal or a group of animals to which a treatment is assigned.

Sampling unit is the fraction or segment of the experimental unit to which a treatment effect is
measured. In most cases in animal production experiment, sampling units represent the entire
experimental unit since the number of the animals to be used in the experiment is most of the time the
entire size of the stock.

2
Treatment refers to the experimental condition or factor that will be imposed on an experimental unit
for the experimental purpose and evaluation. For example, in testing the effect of a new diet in poultry
birds, the different diets fed to the birds are the treatments.

In every experimental design stage, it should be noted that errors emerge as a result of the experiment
itself and such errors are referred to as experimental errors. The sources of these errors can be traced to
2 major reasons:

1. Inherent variation in the experimental units and;

2. Lack of uniformity in the physical conduct of the experiment eg human error as a result of
handling.

Experimental error is therefore the measure of the variation existing among observations on the
experimental units that are treated alike. This is the deviation from the expected value in identically
treated experimental units. This is the actually the first source of variation in most animal production
experiments and it’s the primary basis for testing for significant differences. Experimental errors
emanate from the errors of experimentation, observation and measurement, the variation among
experimental units and the combined effects of all extraneous factors that could influence the behavior
of animals under investigation that have not been considered in the current experiment.

It is natural for experimental errors to occur in animal production experiments but deliberate efforts
must be put in place to reduce to the barest minimum in order not to invalidate the results obtained
from such experiments. To ensure experimental accuracy in animal production experiments, it is
expedient to put in place measures that will reduce experimental errors as follows:

1. Reduce the variation inherent in the experimental materials by using more homogenous
experimental units, replicate your measurements many times to provide additional
information on the experimental units, compare and use information provided by other
variates and grouping the experimental units carefully so that treatments effects are the
major effects on the experimental units.
2. Increase the size of the experiment units through the provision of more replicates or by the
inclusion of additional treatments.
3. Exercise a lot of caution when carrying out the experiment
4. Always ensure an efficient experimental design.

There are actually 3 basic principles in designing a good animal production experiment.

(a) Randomization
Randomization is the process by which the allocation of treatments to experimental units is
done by means of some chance in order to eliminate any form of bias and to ensure that all
treatments receive the same condition without any treatment being favoured or shortchanged.
Allocation of treatments to experimental units is done in the most objective manner such that
all treatments are given equal opportunity of being allocated to any particular experimental
unit. Randomization ensures that the experimental error is a valid and unbiased estimate.

3
(b) Replication
Replication is applying treatment to more than one experimental unit. Another terminology
used in place of replication is repetition. It provides a measure of experimental error needed to
expand the scope of inference of an animal production experiment. It is only when an
experiment is replicated that one can determine whether the differences in the results obtained
from various experimental units are actually due to the effects of the treatments or due to the
inherent variability (experimental errors) among the experimental units. Replicates should
always be independent of one another. The number of replications required in experimentation
varies from one experiment to another depending on the experimental design used. Other
Factors affecting the number of replications include the number of treatments, the magnitude
of the variability in the experimental units, the degree of precision desired and the fund and
time available for the experiment.
(c) Local Control
Local control is the amount of grouping or blocking of the experimental units that is employed in
the adopted experimental design. This entails grouping the experimental units into blocks such
that the units within a block are relatively homogenous while the units between the blocks are
heterogeneous. All the treatments are randomly applied within each block. Local control is also
termed error control since it reduces error variance and thereby makes the experimental design
more efficient.

Procedure for experimentation

Animal production experiments are conducted to have a far-reaching positive impact and
improvements on the methods and means of producing animal products for human
consumption. As such, these experiments must be conducted with utmost care and dexterity to
be able to form the basis for making important decisions in animal production. It is therefore
necessary to plan such experiments very well so as to be able to generate good results from
them. The procedure for experimentation is therefore very important since data acquired
through this process are analysed and interpreted with the respect to the objectives of the
experiment. These procedures are usually determined by the nature of the experiment and the
objective of the study. Such procedure can be summarized as follows:

(a) State the Problem

Animal production experiments are meant to solve particular problems and it is therefore
pertinent from the beginning of all such experiments to properly identify the existing
problem necessitating the research study. The problem should be very clear and focused
throughout the experimentation to ensure the collection and collation of all related and
relevant data.
(b) State the Objective
After the identification of the problem, the objective of carrying out the experiment must be
stated and this objective may be in form of a question to be answered or hypotheses to be
tested or effects to be estimated. The statement of the objective must be clear and

4
 comprehensive enough to capture the experimental units from which the experiment is
conducted.
(c) Design the Experiment
This phase puts into consideration all resources available for the experiment and ensuring
that these resources are well harnessed to meet the research objectives. The choice of an
appropriate experimental design is taken in view of the available experimental materials and
data to be collected. The simplest design should always be employed in cases where more
than one design is feasible.
(d) Carry out/perform the Experiment
The experimental materials are put in place and observations made at this stage. Valid data
are collected at this stage. Care must be taken to ensure that the experiment is conducted
appropriately so that data collected are valid. No amount of statistical manipulation can
produce a good result from a badly conducted experiment. Efforts should be made to
conserve resources such as time, money, personnel and experimental materials without
hindering the scope of the experiment. The data collected should be carefully and correctly
written. Duplicate copies of such records should be maintained and kept in a safe place for
periodic updating as more data emerge.
(e) Analyse the Data and Interprete the Results
The initial experimental design will determine the statistical analysis that would be
employed. All calculations should be carefully checked and crosschecked to forestall error
emanating from calculations. Simple tables and figures should be used to illustrate and
summarize the results of the statistical analysis. Correct decisions should be made
concerning the hypotheses and the results interpreted within the reference of the subject
matter. As long as the experiment has followed all the precautions stated earlier, the
researcher should not exercise restraint in reporting his results even when it seems to be
contradicting previous works.
(f) Prepare the Report of the Experiment
The report of the experiment should be a complete and correct record of the conducted
experiment. It should contain all salient information necessary for another researcher to
repeat the experiment. It should be precise, clear and yet apt without repetition. The report
must meet standards in scheme, arrangement, tables, figures, pictures, nomenclature and
bibliography. These standards are determined by the purpose of the report and the forum
where the report would be presented.

5
Statistical analyses in animal production experiment and the introduction
of computer analyses
There is no doubt that statistical analyses of field experimentation gives meaning and sense into
such research works without which such works would appear meaningless. There are two types
of statistical treatments that could be used for animal production field experimentation viz:
-descriptive statistics
-analytical statistics

Descriptive statistics summarizes the information contained in the data, while the analytical
statistics involves drawing inferences about the population from the data. The analytical
statistics involves the formulation of statistical hypotheses, testing of the hypotheses, and
making inference or drawing conclusions based on the results obtained.

Introduction of computer analyses into animal production experiment

The introduction of the digital technology has relieved many professions of a lot of drudgery and
the hitherto difficult process of calculating statistical analytical procedures has been simplified
by the various computer statistical packages which are being updated routinely to incorporate
modern findings and to be become more user-friendly. These packages are many and include
the Minitab Statistical package, Statwise, Genstat, Statxact and a host of other useful stastistical
packages. These statistical software packages allow you to summarize, display and analyse data.
These can save a great deal of time and trouble. These packages can carry out some or all the
following tasks:
1. Manage data, entered by own spreadsheet or imported from existing computer
files;
2. Illustrate data with graphics such as histograms, boxplots, scatter plots, line graphs,
trellis plots, contour and 3-dimensional surface plots;
3. Summarize and compare data with tabular reports, fitted distributions, and
standard tests, such as t-tests, X2-test and various non parametric tests;
4. Transform data using a general calculation facility with a wide range of
mathematical and statistical functions;
5. Model relationships between variables by linear or nonlinear regression, generalized
linear models, generalized additive models, generalized linear mixed models or
hierarchical generalized linear models;
6. Design and analyse experiments, from one-way analysis of variance to complex
multi-stratum designs, using a balanced-ANOVA or a REML approach (including the
modeling of correlation structure);
7. Identify patterns in data by means of multivariate techniques such as canonical
variates analysis, principal components analysis, principal coordinates analysis,
correspondence analysis, partial least squares, classification trees and cluster
analysis;
8. Analyse results from stratified or from unstructured surveys;

6
 9. Plot control charts, print Pareto tables and calculate capability statistics;
10. Analyse time series, using Box-Jenkins models or spectral analysis;
11. Analyse repeated measurements, by analysis of variance, or using ante dependence
structure, or by modeling the correlation over time;
12. Analyse spatial patterns, using Kriging or spatial point processes.

Let us now use a particular statistical package to solve a simple Completely Randomized Designed
experiment.