NEW CONCEPT OF CHEMOMETRICS

Abstract

The parametric statistical methods of the research experimental data most often use presently in chemometrics. If experimental data do not correspond to the normal probabilistic distribution that in this case it is impossible prodice adequate data processing. At recently in chemometrics more often started to use the classical nonparametric methods. The classical nonparametric methods do not handle the raw datas probabilistic distribution. But these methods when undertakind the real calculations don’t care use that or other types of the distributions. The classical nonparametric methods under its realization usually requare referencing to corresponding statistical tables. Its own table is used for each such method. Chemometrics to presently presents itself many not bound between itself statistical methods of the investigations. The huge defect of the classical methods of chemometrics (parametric and nonparametric) in absence of the united approach to processing the raw datas. In this article is offered in fundamentally new approach to statistical investigations of the datas. If use the method of nonparametric bootstrap that possible replace many unbound between itself methods of classical chemometrics whole one ore two universal methods. And these universal methods have not what or essential defect. With standpoint of the program realization this means presence one or two universal procedures for decision nearly all practical problems of chemometrics. One of the primary tasks of nonparametric bootstrap as follows problem of the duplication of the sample is considered in article. Herewith simulated that was organized not one series of experiment but well over (for example 1000 … 10000). On example is shown use the method of nonparametric bootstrap with finding confidential interval for average and median for ecological time series. On programming language MATLAB is brought code corresponding bootstrap procedure.