Estudo de metodos multivariados para analise e calibração de espectros

AUTOR(ES)
DATA DE PUBLICAÇÃO

2003

RESUMO

Considering the importance of resolving the chemical analysis mixture problem, a comparative study involving R-mode principal component methods, principal component regression (PCR) and partial least squares (PLS) regression, and the Q-mode principal component method of Imbrie was carried out. Furthermore the K matrix method was also studied. Several pre-processing methods were used for the different methods. Cross-validation and percentage variance vs. number of principal component graphs were used to determine the number of mixture components. To evaluate the different methods (PCR, PLS and Q-mode PCA of Imbrie) all were applied to the infrared spectra of isomeric mixtures of ortho, meta and para-xylene, silica-alumina-mulite mixtures and drug mixtures containing cocaine, phenacethine, lydocaine and benzocaine. Model results were evaluated using standard errors of prediction (SEP), root mean square errors of prediction (RMSEP), F tests and linear correlation coefficients. The Q-mode principal component method provided predictions in better agreement with the proportions that were used to make the mixtures or that were known to constitute the mixtures. It was possible to recuperate the spectra of each of the pure mixture constituents using the K matrix method. In another part of this work the possibility of using the Q-mode method to determine dissociation constants of 2,7-naphtalenediol (2,7-ND) using ultra-violet-visible region spectra was investigated. The dissociation constants determined were close to the literature values showing that Q-mode principal component analysis is especially useful to determine constituent proportions in the absence of experimental calibration sets.

ASSUNTO(S)

calibração multivariada analise de componentes principais minimos quadrados

ACESSO AO ARTIGO

Documentos Relacionados