Copyright © 2014
Data Analysis Australia

Shoemakers' Lasts and Model Ships

What do shoemakers' lasts and model ships have in common?  The answer is that they are both models - in each case while the model is only an approximation to something, the approximation is good enough to be useful.  A last is an approximation to a foot and although it looks like a lump of wood, it is useful in making a shoe that fits well.   A model ship is a fraction of the size of the real thing but can be used to understand how it moves through the water.

While Data Analysis Australia is not in the habit of building model ships, the same principles of simplification and utility are relevant to the models we produce.  Whether they be conceptual models to better understand or simplify a process, or detailed data driven statistical or mathematical models to provide a theoretical understanding of relationships in a quantifiable way, they need to be formed in a way that captures the important characteristics while simplifying the superfluous.

There are many software packages available these days that include modelling capabilities.  These tools allow professionals in any field to create models. However, it is not the simple act of using the software to produce a model that is the critical part of a model's development - it is the verification, confidence and knowing that the right model has been chosen that makes the task one for the experts. 

The skill is in knowing how far a model can be simplified while still including enough information to adequately represent the real world situation.  This requires experience not only in the knowledge of the situation but also about the models and data structures themselves.  

Data Analysis Australia sees modelling as being more than just process driven and more than just data driven.  Our holistic approach is to first understand the real life process that is being modelled and to then use data to design and calibrate relevant statistical and mathematical models to apply to the process.  Without this duality of understanding the resulting models could be meaningless, or even wrong.

Data Analysis Australia has implemented this duality for many clients.  The forecasting of future court requirements required us to first obtain a detailed understanding of the courts' processes, then calibrate these models with statistical and mathematical estimates of crime rates, population forecasts and average hearing lengths.  

In modelling the performance of both a current method and a proposed new method of processing ore, an understanding of the processes enabled Data Analysis Australia to model the performance of the two methods upon numerous characteristics of the ore, assisting the mining company in further refining its new methods.

A thorough understanding, not only of the exogenous factors that affect water consumption, but also the statistical properties of the data, is required to measure the effect of advertising campaigns on the reduction of water consumption.  Data Analysis Australia used sophisticated generalised linear models to quantifiably, robustly and accurately measure the effects of these factors.  

This duality of understanding gives Data Analysis Australia the confidence it requires to present relevant, reliable results and recommendations to the client.

For further information regarding Data Analysis Australia's modelling expertise, contact Anna Munday