Nils B. Vogt

Quality by design. Managing research and development

Abstract Vogt, N.B., 1992. Quality by design. Managing research and development. Chemometrics and Intelligent Laboratory Systems , 14: 93–101. Quantitative methods of analyses and prediction which make multivariate and complex data more comprehendible contribute to make research and development (R&D) management decisions more objective. The scientific fields of * metrics, e.g. chemometrics, technometrics or qualimetrics, encompass powerful tools for achieving technological competitiveness. The paper describes some ideas that have evolved through considering the possibilities of using * metrics in the management and technological issues of the research and development and design phase of industrial projects and suggests how these fields may also contribute to a common understanding of the relationships between the business, strategic R&D and operational units in a company.

Multivariate classification of histochemically stained human skeletal muscle fibres by the SIMCA method

SummaryThe SIMCA (soft independent modelling of class analogy) method of pattern recognition has been used to classify four muscle fibre types: I, IIA, IIB and IIC. The samples were histochemically stained human skeletal sections from biopsy material. Disjoint (separate) class modelling gave information about variables, i.e., the combinations of alkaline, acidic and Ca2+-containing preincubation procedures with appropriate discrimination power, and showed satisfactory separation of the classes (fibre types). Two serial stained muscle sections represent a minimum for a proper classification of the four fibre groups. A comparison of biopsy samples from two different persons showed significant variation in the data structure between similar fibre types, probably caused by intermuscle variations. It is suggested that the introduction of computer-assisted classification by the application of such multivariate analytical techniques both facilitates the classification of muscle fibres and improves the precision and reliability of fibre typing.

Environmental Applications of Chemometrics - Envirometrics

Pollution in the environment is measured as a combination of changes in chemical concentrations, physical conditions and biological systems. By using multivariate methods on the data obtained from extencive measurement programs the environmental analyst attempts to accomplish three objectives; Phenomena-description (representivity), Dimension-reduction (interpretability), and increased Resolution (sensitivity). Three different problems are used to illustrate the use of multivariate methods in envirometrics. The three applications are: 1) Multivariate correlation patterns of alkyl-homologue series, 2) Composition Activity Relationships (CARE) and 3) Multivariate Ecotoxicological Mapping (MEM).

Evaluation of the direct and autocorrelated use of physicochemical descriptors in quantitative structure-property relationship studies. A partial least squares prediction of log P for chlorinated alkylbenzenes

Abstract Bye, E., Heiberg, A.B. and Vogt, N.B., 1990. Evaluation of the direct and autocorrelated use of physicochemical descriptors in quantitative structure—property relationship studies. A partial least squares prediction of log P for chlorinated alkylbenzenes. Chemometrics and Intelligent Laboratory Systems, 9: 185–199. The use of global and substituent physicochemical structure descriptors for quantitative structure—property relationship (QSPR) studies has been evaluated through direct and autocorrelated applications. Partial least squares (PLS) analysis has been used to predict the partition coefficient for octanol/water, log P, for chlorinated alkylbenzenes (CABs). This investigation revealed that various of the physicochemical parameters considered were highly correlated with the partition coefficient and that PLS models based on global, direct or autocorrelated substituent parameters all seem to be adequate for prediction of log P, as compared to results obtained experimentally or by means of standard substituent calculation schemes. However, the use of autocorrelation vectors of connectivity, steric and electronic effects may be preferably due to a high degree of explanation of the molecular descriptor variance.