Defra - Department for Environment, Food and Rural Affairs.

Science Search

Science and Research Projects

Return to Science Search homepage   Return to Project List

Modern statistical methods in analytical chemistry. - FS2917

Objective
Among the techniques to be investigated are the following: (I) Robust statistics constitutes a modern approach to the interpretation of data that include outliers. The limited number of applicants so far reported in analytical science have been uniformly successful. The applicants have already made a start in this direction having contributed a substantial proportion of the existing literature on the subject. (ii) Computer-intensive methods, including the bootstrap and crossvalidation which are sample-based methods for estimating variances and confidence intervals of complex statistics, have been used successfully in analytical chemistry only within the last year, largely by the applicants. Although the theory of these methods is complex, the application is straightforward and may well be suitable for general use. (iii) Effective visual presentation of data is often as important as statistics, and it is important to exploit modern computer capabilities to generate novel approaches. (iv) Bayesian statistics has been virtually ignored in analytical circles, despite several potentially valuable applications. Bayesian statistics provides a formal framework for the combination of prior information, whether objective or subjective, with data. (v) Range data. The procedures used in the validation of analytical methods and results require good design followed by the collection and statistical interpretation of data. Procedures such as collaborative trials, proficiency tests and internal quality control fall into this category. Data produced by test kits is increasingly being studied in this context as analysis moves into the workplace. However test kits often produce data in the form of ranges only, in the most extreme case being dichotomised into present/absent. Standard statistical methods do not apply in such cases. In fact there seems to have been very little research into how to deal with a whole range of issues concerned with the evaluation of oligochotomous (range) data, and there is urgent need for suitable statistical methodology to be identified and developed. This has been identified as an important issue by the Statistics Subcommittee.
Time-Scale and Cost
From: 1997

To: 1999

Cost: £86,718
Contractor / Funded Organisations
Royal Society of Chemistry
Keywords