Modelling component combinations by means of attention function scores

• Published in: Analytica chimica acta 1983, Vol.150 (1), p.207-217
• Main Authors: Goldschmidt, Henk M.J., Leijten, Jan F., Scholten, Marc N.M.
• Format: Article
• Language: English
• Subjects: Biological and medical sciences; Fundamental and applied biological sciences. Psychology; General aspects; Mathematics in biology. Statistical analysis. Models. Metrology. Data processing in biology (general aspects)
• ISSN: 0003-2670; 1873-4324
• DOI: 10.1016/S0003-2670(00)85472-0

The determination of a large number of components in a small sample is common practice in clinical chemistry. The optimal combination of components to be assayed in connection with a particular problem can be difficult to establish. This paper describes an attempt to achieve an optimal combination, not by means of the correlation matrix of component concentrations but by quantification of the attention that is given to the result for each component by the person requesting the analysis. Here, attention is defined as an intention to take the action. It is suggested that little attention is paid when the result is within or near certain expected limits, while a growing deviation from these limits will result in increasing attention. More attention will be paid to a larger deviation, resulting in ultimate saturation. This reasoning results in a sigmoidal curve on each side of the distribution. Missing data are no longer a problem, as they get zero attention. The actual curves, called attention functions, were established with the aid of the persons requesting the analyses, and with a derivation from the cumulative frequency distributions of the component concentrations. A set of 29 attention function scores was collected for each of 298 samples. A hierarchal cluster analysis was applied to these attention function scores to discover component similarities with regard to the attention eventually given to them. The combinations of components found were readily understandable. The advantage of this approach is that the criteria for combining components is directly linked to the daily practice of interpretation of the component concentrations by the people submitting the samples.