Scribd
Upload
84 views2 pages

SST 412 Design and Analysis of Experiments

This document provides definitions and principles for the design and analysis of experiments. It defines key terms like experimental unit, factors, levels, treatments, response, blocks, plots, randomization, replication, and local control. Randomization involves allocating treatments to plots randomly to guarantee observations are random variables. Replication means each treatment appears more than once to minimize error in estimating means. Local control means including a control treatment to compare other treatments against and help minimize observational error. The overall goal is to select a field plan that minimizes error.

Uploaded by

Esekon Geoffrey Lokol
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
84 views2 pages

SST 412 Design and Analysis of Experiments

This document provides definitions and principles for the design and analysis of experiments. It defines key terms like experimental unit, factors, levels, treatments, response, blocks, plots, randomization, replication, and local control. Randomization involves allocating treatments to plots randomly to guarantee observations are random variables. Replication means each treatment appears more than once to minimize error in estimating means. Local control means including a control treatment to compare other treatments against and help minimize observational error. The overall goal is to select a field plan that minimizes error.

Uploaded by

Esekon Geoffrey Lokol
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 2

SST 412 DESIGN AND ANALYSIS OF EXPERIMENTS

CHAPTER 0: INTRODUCTIOMN

The subject of Design of Experiments is concerned with:


Planning an experiment
Making relevant observations from the experiment
Analyzing results of the experiment

Definition 1: Experimental unit: Consider a large farm. suppose we want to try out a set of
fertilizers. Then we have to divide the farm into convenient small units on which the
experiment can be carried out. These units are called experimental units.

Definition 2: A factor is a regressor variable e.g. wheat type, sulphur concentration etc
Quantitative factor: For this the levels are numerical e.g. 300lb/acre, 600 lb/acre etc

Qualitative factor: For this level are not numerical e.g. what type B, Method A of teaching
etc

Definition 3: A level of a factor is a particular form of that factor e.g. wheat type B, sulphur
concentration(e.g. 300 lb/acre).
Definition 4: Treatments: Various objects of comparison in a comparative experiment are
known as treatments e.g. fertilizers, crop varieties, different methods of cultivation etc

Definition 5: For a single factor study treatment is one level of that factor. For a multifactor
study a treatment is one combination of the factor levels e.g. if we have two factors: Calcium
at levels 2, 4, 6 and Potassium at levels 1,8 ,10, then a combination (Calcium,
Potassium)=(4, 10), say, will be an example of a treatment in a 2-factor study.

Definition 6: Yield/Response -the effect of a treatment

Blocks and Plots: An experimental unit can be divided further into subunits called blocks.
Each block can also be divided into smaller units called plots.

Principles of Experimentation:
According to R.A Fisher, the basic principles of design of experiments are:

1. Randomization
2. Replication
3. Local Control
Randomization
Consider an experiment designed to test productivity of ‘v’ treatments The ‘v’ treatments
could be ‘v’ varieties of maize. These ‘v’ treatments or varieties are allocated to the plots in
the field at random or in a random order which in turn guarantees that the observations
coming from any plot is a random variable. This assignment of treatment to the plots is
known as Randomization

Replication
It is necessary that each treatment should appear more than once so as to minimize the error
in estimation. For example suppose treatment ‘ i ’ occurs in ‘ ni ’ plots with observations
yi1 , yi 2 , ..., yij , ..., yini each with error equal to  (unknown) then the mean

1
ni

y i. 
ni
y
j 1
ij has error equal to
ni


Replication is done in order to


a. estimate the experimental error i.e. error due to random variation
b. reduce the experimental error of the mean yield

Local Control
In planning an experiment there should be a minimum of two treatments. One treatment can
be regarded as control treatment and the performance of the other treatment can be compared
with the control treatment.

The main objective of any field plan is to minimize the error in observation  . We select that
field plan for which  is smallest. One particular field plan is known as Completely
Randomized Design(CRD)

You might also like