Premium
Using properties of random matrices for target factor analysis of sensor array data
Author(s) -
Marth M.,
Maier D.,
Honerkamp J.,
Rapp M.
Publication year - 1998
Publication title -
journal of chemometrics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.47
H-Index - 92
eISSN - 1099-128X
pISSN - 0886-9383
DOI - 10.1002/(sici)1099-128x(199807/08)12:4<249::aid-cem512>3.0.co;2-g
Subject(s) - eigenvalues and eigenvectors , matrix (chemical analysis) , calibration , data matrix , algorithm , covariance matrix , mathematics , statistics , computer science , chemistry , physics , chromatography , clade , biochemistry , quantum mechanics , gene , phylogenetic tree
Target factor analysis is an important issue in the analysis of sensor array data as it allows one to test whether measurements contain only the substances with which a chemical sensor system was calibrated. In this paper a new approach based on the properties of random matrices is presented. The problem is first transformed to a pseudorank estimation problem by forming a combined calibration – prediction data matrix. Then the largest eigenvalue of the estimated measurement error matrix of this matrix is compared with maximum values obtained from pure random matrices. The test is statistically exact and especially useful for sensor array data. The largest eigenvalue test is compared with Malinowski’s F‐test on simulated data and tested on real data from chemical sensor arrays. © 1998 John Wiley & Sons, Ltd.