2003 IRISH SCIENTIST YEAR BOOK

Home Page

Table of contents

Index by Author

Index by topics

Search


Institute of Technology Carlow
Norman McMillan, Greg Doyle & Martina O'Neill
Democracy is not always the best method or rule!

Holmium UV-visible spectrophotometer spectra and resulting vertical data-scatter revealing that the instrumental reproducibility error is only in the photometric (vertical) response of the instrument under test.

A new approach to data mining is being developed at the Institute of Technology Carlow. The method produces a visualisation of the differences in two digitised data sets. The data sets are compared point-by-point by an optimised algorithm. The algorithm usually treats all points in an egalitarian or non-democratic way. The approach can however weight the 'voting' of the respective data point pairs in a way that helps elucidate a specific problem.

An example best illustrates the approach - The 'performance quality' of UV-visible spectrophotometers is established using Starna NIST (National Institute for Standards and Technology) standards. The wavelength is typically checked using a Holmium standard giving a spectrum with very sharp peaks occurring at wavelengths that are very accurately known (see Figure). It should be obvious from inspection of the data-scatter plot, produced by the analysis of two consecutive measurements of this standard, that an excellent visualisation of its performance has been obtained. In the case shown, the photometric reproducibility of the instrument is poor.

The advantage of the technique over the traditional is that it analyses the entire data set and all points here are scattered only in the photometric direction. It is thus visually obvious that the instrument has negligible wavelength errors and only measurable photometric errors. For instruments with wavelength reproducibility errors, the scatter would be expected to be in the opposite direction. For the instrument that has reproducibility errors in both its wavelength and photometric operation, loops are observed.

The Carlow software gives a numerical value for the average scatter of all the data points from the projection radius measurement, and secondly it provides a coherence angle measure, which allows the user to see immediately which of the two instrumental errors is the greater. If the angle is 0 degrees, then the data differ only in that there is wavelength scatter; if this angle is 90 degrees, then there are only photometric errors. It is important to point out that this technique can be used to monitor the performance of ANY instrument that produces a graphical output.

The simplest example of a non-democratic algorithmic approach is in the treatment of noise, which scatters into a central area beyond a specific projection radius determined by the root-mean-square noise value. Random noise scatter has been shown to follow a specific algebraic form in the data-scatter transform from a signal in a way that can be controlled to any level.

Contact: Dr Norman McMillan; Tel: 059 91 70554; E-mail: [email protected]