Model building is best approached as a scientific exercise. The goal is to build a case for taking some action such as exploring how the system behaves further with new experiments or defining how a process or product will be configured.
Models are ways of interpreting the evidence for deciding how to act, and as usual it’s good practice before making a decision to consider the strength of the evidence and possible alternatives. Finding that the model doesn’t fit the data well or violates the underlying assumptions undermines it as evidence for a decision, so typically in this situation you try to rectify these issues by changing the model in some way (a matter of manipulating the form of the model and the raw data it is built from) to see if this suggests a significantly different course of action.
In many cases the decision does not materially change, in which case you now have better evidence in its favour. In others a different course would be suggested and you may need to either make a choice between the alternatives based on prior knowledge or other factors, or ideally do more work to gather evidence on each side.
The documents in this section introduce the main tools for this process: transformations and row sets.
A background document introducing transformations and row sets, how to use them and why they work.
Technical background explaining transformations with reference material on implementation in Synthace.
Instructional document explaining how to use row sets.
Instructional document explaining using predefined transformations.
Instructional document explaining how to do column-based calculations for data preparation
Instructional document explaining how to use the generic Python interface to transform data.