Linear Statistical Modelling is a set of tools and techniques used to understand how a system behaves in quantitative terms. By making some fairly simple assumptions about how the system operates in general these tools can be used to define a detailed model of how changes to a set of input variables presumed to influence the system’s behaviour result in behavioural changes as represented by one or more output measurements. DOE typically makes use of this kind of modelling as it is sufficiently powerful and flexible to work on a very broad range of systems while being simple and tractable enough that designs can be calculated in seconds or even by hand.
Linear modelling is a very broad topic with many possible applications. For the sake of clarity, brevity and relevance, we will focus here on what is necessary and important for scientists doing DOE in the Synthace platform. This means a great part of otherwise interesting and useful material had to be omitted. Similarly, you will not see a deep dive into the underlying mathematical concepts, but you will find a few references for further reading at various points if (as we hope!) your interest is sufficiently piqued to delve more deeply.
Model building is a craft in its own right - almost as much art as science - and one which really only comes with experience. So don’t expect to understand everything just by reading, to really see how it all works you need to go out and try it for yourself.
In the following documents, you will be gradually introduced to the concepts and techniques of linear statistical modelling for DOE analysis purposes.
A general introduction to modelling for DOE including background information.
An introduction to the fundamental structure of linear models.
A brief overview of how the modelling process works.