Chaughule, Thorat - Statistical Analysis/Design of Experiments
Function of Statistics in Experimentation
Planning Phase
- Problem recognition and formulation.
- Selection of response or quality characteristic.
- Selection of process variables or design parame-
ters. - Classification of process variables.
- Determining the levels of process variables.
- List all the interactions of interest.
Design Phase - Control known sources of variation
- Allow estimation of the size of the uncontrolled vari-
ation - Permit an investigation of suitable models
Analysis Phase - Make inferences on design factors
- Determine the design parameters or process va-
riables that affect the mean process performance. - Determine the design parameters or process va-
riables that influence performance variability. - Determine the design parameter levels that yield
the optimum performance. - Determine whether further improvement is possi-
ble. - Guide subsequent designs
- Suggest more appropriate models
8.3.1. Planning Phase
Statistical considerations should be included in the project planning & designing
phase of any experiment. At this stage of a project one should consider the nature of the
data to be collected, including what measurements are to be taken, what is known about
the likely variation to be encountered, and what factors might influence the variation in
the measurements.
8.3.2. Design Phase
In this phase, one may select the most appropriate design for the experiment. Expe-
riments can be statistically designed using classical approach advocated by Fisher, or-
thogonal array approach advocated by Taguchi or variables search approach promoted
by Dorian Shainin. In Fisher approach, choice of full factorial, fractional factorial or
screening designs such as Plackett-Burmann design can be used (Antony, 2003). The
size of the experiment is dependent on the number of factors and/or interactions to be
studied, the number of levels of each factor, budget and resources allocated for carrying
out the experiment, etc. During the design stage, it is quite important to consider the
confounding structure and resolution of the design.
A statistical design should be selected that controls variation from known sources.
The design should allow the estimation of the magnitude of uncontrollable variation and
the modeling of relationships between the measurements of interest and factors
(sources) believed to influence these measurements. Uncontrollable variation can arise