All programmes > Experimental Design

Experimental Design (B-KUL-G0B68A)

4 ECTSEnglish25 Second termCannot be taken as part of an examination contract
Goos Peter |  Strouwen Arno (substitute)
POC Master in statistiek

Aims

This course deals with modern methods for setting up highly informative experiments. More specifically, it provides an in-depth treatment of the optimal experimental design approach, which is extremely flexible and can handle all kinds of practical constraints that may occur in the planning phase of an experiment. During the course, the students learn how to use the JMP software, which is state of the art for the optimal design of experiments. The focus is on factorial experiments, i.e., experiments studying multiple treatment factors.

Previous knowledge

Students should have good knowledge about linear algebra (simple matrix operations, inverses), the basic principles of probability and statistics, and about the basics of regression and analysis of variance (or about linear models in general). They should be familiar with the concepts of confidence intervals, hypothesis testing, p-values, power calculations, the fitting of simple linear models and the interpretation of fitting diagnostics. 
 
Prerequisites: Students should have had an introductory statistics course and a course covering the basics of regression and analysis of variance. The course “Linear Models: Regression Analysis and Analysis of Variance” is sufficient as a prerequisite (although other course(s) may also be sufficient).

Identical courses

This course is identical to the following courses:
G0B68B : Experimental Design

Is included in these courses of study

Activities

4 ects. Experimental Design (B-KUL-G0B68a)

4 ECTSEnglishFormat: Lecture25 Second term
Goos Peter |  Strouwen Arno (substitute)
POC Master in statistiek

Content

This course discusses the design of factorial experiments. Initially, the focus is on completely randomised experimental designs. Next, the focus shifts to experimental designs involving a restricted randomisation. First, the concept of blocking is discussed. Next, split-plot and strip-plot designs are studied.

The emphasis in the course is on the optimal design of experiments. In optimal design of experiments, the experimental design is tailored to the problem at hand (unlike classical experimental design, where inflexible, standard designs are chosen from catalogs). The course builds on concepts from regression and analysis of variance, such as fixed and random effects, power calculations, variance inflation factors, multicollinearity, confidence intervals, prediction and lack-of-fit tests.

Every topic in the course is introduced and illustrated by means of a case study from industry. The case studies are realistic in the sense that they involve quantitative and qualitative experimental factors, experimenters have to deal with limited budgets and difficulties to randomise, and forbidden combinations of factor levels. In each of the case studies, the goal is to enhance to performance of a process or a product. The areas of application are the food industry, the pharmaceutical sector, and the metal and chemical industries, among others.

The statistical software package used is JMP.

Course material

The textbook used is "Optimal design of experiments: A case study approach", co-authored by Peter Goos and Bradley Jones. A hard-copy of the textbook can be brought to the exam (photocopies, prints and e-books are not allowed at the exam).

Format: more information

The students are expected to master the concepts and understand the properties of and the differences between the different optimal experimental designs by studying the material from the lectures and the textbook. By means of an assignment and two PC sessions, the students get the opportunity to familiarise themselves with the JMP software required to plan experiments optimally.The assignment also requires the students to conduct a virtual experiment, so they can put the acquired knowledge into practice.

Is also included in other courses

Evaluation

Evaluation: Experimental Design (B-KUL-G2B68a)

Type : Partial or continuous assessment with (final) exam during the examination period
Type of questions : Open questions
Learning material : Course material, Calculator

Explanation


The exam in June counts for 15 of the 20 points for the course. The assignment counts for 5 of the 20 points in June. A hard copy of the book "Optimal Design of Experiments: A Case Study Approach" can be brought to the exam (photocopies and prints of e-versions of the book are not allowed).

The details and consequences for students who do not submit the assignment (or do not submit the assignment in time) can be found on Toledo.

Information about retaking exams

The exam in August/September counts for 20 points. The score for the assignment is not carried over to August/September. A hard copy of the book "Optimal Design of Experiments: A Case Study Approach" can be brought to the exam (photocopies and prints of e-versions of the book are not allowed).