Design of experiment.

Language

French
> English-friendly

Prerequisites

Concepts and tools equivalent to those taught in teaching units

LSTAT2120	Linear models
LSTAT2020	Logiciels et programmation statistique de base

Main themes

- Experimental cycle and strategies - Linear regression as a tool to analyse the results of a designed experiment - Problem formalisation and qualities of an experimental design - Factorial designs and derivatives - Designs for the estimation of response surfaces - Optimal designs - Experimental design as viewed by Taguchi - Designs for mixture experiments - Simultaneous optimisation of several responses - Simplex and EVOP methodology to optimise one response

Learning outcomes

At the end of this learning unit, the student is able to :
1	At the end of the course, the student will be awared of the interest of using a methodology to design experiments that provides a maximum information at the lower cost. He will gain knowledge on different possible classes of experimental designs and on the statistical methods available to analyse experiment results.

Content

The themes discussed in this course are :

• Experimental cycle and strategies
• Linear regression as a tool to analyze the results of a designed experiment  
• Simultaneous optimization of several responses
• Problem formalization and qualities of an experimental design
• Screening designs
• Factorial designs and derivatives
• Designs for the estimation of response surfaces
• Optimal designs
• Designs for mixture experiments
• Blocking.
• Designs for the estimation of variance components.  

The course is organized into modules, a subset of which constitutes partim A, which can be taken for 3 credits instead of 5.

Teaching methods

Lectures (30h)

• Methods presentation on the basis of real-life situations.
• Formal but intuitive discussion of theoretical concepts and formulae for most methods.
• Interpretation of software outputs and use of appropriate softwares in class.
• Interactive lectures: students are encouraged to participate during the course.

 Computer labs (10h)

• Case studies on JMP, methodological exercises, and JMP Output interpretation. 

Homework and projects

• The student is invited to prepare each week an exercise, a quiz or a small project in order to apply and integrate course content.  

Evaluation methods

The course grade is based on :

1. Continuous assessment activities
   1. Compulsory tasks during the term.
   2. Online quizzes during the term.
2. A written exam on the course content (“theory” and methodological exercises).
3. Applied project (in groups of 2 students) leading to a written report and oral presentation followed by questions.
4. A practical exercise test in the computer room.

For students taking the full course,

• continuous assessment counts for 10% of points
• the written exam accounts for 60% of points
• the project counts for 30% of points
• the computer room test is a non-supervised (and optional) self-study activity.

For students taking Partim A only,

• continuous assessment accounts for 10% of points,
• the written exam accounts for 70% of points,
• The computer room test accounts for 20% of points.
• the project is optional.

In the overall grade, points for continuous assessment, the project or the computer room test are only acquired if the student has obtained a grade of at least 45% in the written exam, otherwise the grade for the written exam replaces the overall grade.

Other information

Prerequisites:

• Basic training in probability and statistics: descriptive statistics, basic statistical inference, multiple linear regression.
• Ability to use a personal computer fluently: file handling, use of Word and Excel.

Documentation

• Everything is available on the moodle website

Software
Several experimental design software packages are available in the training room and for download to personal computers.

Online resources

See the Moodle site for more information

Bibliography

• Box G. et Draper N. et H. Smith [1987], Empirical Model-Building and Response Surfaces, Wiley, New York
• Khuri A. et Cornell J., [1996], Response surfaces : designs and analyses, Marcel Dekker.
• Myers R.H., Douglas C. Montgomery [2002], Response Surface Methodology: Process and Product Optimization Using Designed Experiments. Wiley
• Et beaucoup d'autres possibles...

Teaching materials

• Voir le site Moodle

Faculty or entity

> LSBA