Experimental Design
General context
The mission of this course is to explain the underlying principles and concepts of experimental design what will allow the course takers to understand why well-designed experiments are more efficient. We will apply these concepts and use R tools to design or improve concrete and typical biotech research experiments.
We welcome experimental design questions from the audience.
This training is organised in collaboration with the Helis Academy. More information see https://helisacademy.com/
Researchers with some experience in designing experiments, but that would like to better understand the statistical principles behind it with the aim to improve the relevance and efficiency of their experiments.
A basic knowledge of R is required to follow this course.
- This course if free for VIB participants
- Non-VIB participants pay a fee of 100 euro
- This course is not open for industry
- Note that upon no-show without valid justification you will be blacklisted for the VIB training program for 1 year and a fee of €100 will be charged. Click here for more information
Trainers
Joris De Wolf
Joris De Wolf has a long track record in biostatistics. He worked at Crop-Design, later BASF, as team leader of the biostatistics group. He was involved in the experimental design and start-up of the high-throughput phenotyping system TraitMill, design of the databases and streamlining of statistical analyses. Furthermore, he was responsible for the field testing pipeline and statistics on transgenic yield improvement. In 2016, he joined GSK vaccines. Currently he works full time as a biostatistics consultant.
Program
- Introduction
- Why do we do experiments?
- Difference between experiments and observations
- Causation and correlation
- Think twice before you start: importance of careful design
- Get the reseach question clear
- Context and inference space
- Desired and undesired sources of variability
- Selecting the measurements
- Plenty rough or a few accurate observations
- Important concepts:
- Orthogonality
- Randomization and blocking
- Replication and power
- Independence
- Fixed and random effects
- First simple case: a design to determine a regression
- Power studies
- Intro
- Hands-on in R via traditional methods and via simulations
- Apply it to the regression problem
- One-factor-at-a-time vs factorial designs
- Interactions
- Factor screening vs optimisation
- Fractional designs (standard designs, optimal designs)
- Finding good design and power study for factorial experiment with R
- Randomisation schemes
- Completely randomized design
- Complete block design
- Incomplete block design
- Splitplot design
- Case studies
- Group discussion of existing case
- Group work: design best experiment based on available material and info
- Recapitulate
- Closure