Generalized PLS Discriminant Analysis for the “multimodules” satisfaction evaluations

The knowledge of the customer evaluation of a given service represents a competitive advantage in order to orient the renewal of company offer as well as the pricing itself. In literature several methods have been proposed for the service quality assessment. A lot of these models (Servqual, Normed Quality, Qualitometro etc.) are based on the Gap Theory of Service Quality, developed for the profit sector by Parasuraman et al. (1994). These models measure the gap between customer's expectations for excellence and their perceptions of actual service delivered. Servqual undertakes to measure service quality across five dimensions: reliability, assurance, tangibles, empathy and responsiveness. Several multidimensional statistical methods have been proposed in literature for the description and the exploratory study of these data. Aims of these proposals are to estimate the multidimensional aspects of the investigated system and the introduction of criteria of judgment (D’Ambra et al. (1999); D’Ambra, Amenta (2000a); D’Ambra, Amenta (2000b). Deeper analysis can be considered when we have questionnaires submitted to several groups or taken in different times: for example, in the student satisfaction, we have questionnaires taken in different K modules or times (for example, during a post-degree training course with different teachers at the beginning and the end of each module) or questionnaires, in the patient satisfaction, submitted in a survey in different wards. In these cases, often with few questionnaires, we have to deal with the study of groups (or modules) and with the research of their separation with the aim to fix a decision role. In literature several methods have been proposed to the problem of discrimination (To- massone et al., 1988; Saporta, 1990). In some situations, a misuse of PLS Discriminant Analysis (PLS-DA) (Sjöström et al., 1986) can lead to a biased solution which does not answer the given problem of discrimination. This unsuitable solution is due to several reasons (Sabatier et al., 2003). In a patient satisfaction framework, aims of this paper is to consider an extension of the generalization of PLS (GPLS) proposed by Cazes (1997) but for the context of discrimination (Sabatier et al., 2003). A patient satisfaction study is con- sidered. The results will be presented without reference to all the reports and graphical outputs provided by PLS method to make the paper more readable.