| چکیده انگلیسی مقاله |
The use of multivariate methods in order to analyze chemical data sets and extract useful information is inevitable. Multivariate curve resolution is the popular resolution methods. But, the results are always associated with the rotational-ambiguity (RA) problem. By using chemical information in the form of constraints, the amount of rotation can be reduced. Since applying different physical/chemical constraints can effect on the extent of RA and the accuracy of the SMCR results, herein, the effects of presence and utilizing selectivity constraint on the accuracy of the results is investigated. The selectivity constraint was defined as regions with presence of one component in the concentration or spectral profiles [1]. It is assumed that in the local rank-based methods the number of components in each region is equal to its rank. Thus, implementing local rank information as a constraint in can significantly diminish rotational ambiguity. It has been showed that the implementing of local-rank as a mathematical constraint for restricting the rotational ambiguity can lead to incorrect solutions due to the local-rank-deficiency problem [2]. Thus, obtaining information about selective windows by local rank information is not reliable. In order to investigate the effect of presence, detection and applying the information of selective regions, different simulated data and real experimental was checked out (hyperspectral image and HPLC-DAD). The effects of presence of selective window(s) of profiles on accuracy of the MCR methods was investigated. For this, the range of feasible solutions by using only non-negativity constraint was calculated. Then, the selectivity of resolved profiles for rank-one windows, for different solutions were checked out. Presence of one, two, or three components can be observed in sub-windows with rank one. Thus, due to the rotational ambiguity, there is an ambiguity in the presence pattern of concentration profiles. Thus, the local-rank-one sub-windows is not necessarily selective. This is the local-rank deficiency problem which is unavoidable in local rank-one sub-windows. It is notable that, although different sets of MCR solutions have variety of presence pattern, all are consistent with the scheme of rank windows obtained by EFA. The selective window size also varies in different feasible solutions. When the data conditions are such that the profile of at least one of the involved components is unique, then, because there is no rotational ambiguity for that profile, the observed selectivity status is true. Thus, it is possible to detect real selective profiles in data only with non-negativity constraint. |
