Department of Management

PLS and multicollinearity under conditions common in satisfaction studies

Research output: Contribution to conferencePaperResearchpeer-review

  • Department of Marketing and Statistics
A number of studies have investigated the performance of the PLS path modelling algorithm in the presence of common empirical problems, such as model misspecification, skewness of manifest variables, missing values, and multicollinearity, and they have shown PLS to be quite robust (see e.g. Cassel et al., 1999; Kristensen, Eskildsen, 2005). However, most of the studies, including our own, have focused on somewhat simple models with very simple correlation structures. This paper extends the existing knowledge by investigating the effect of varying degrees of multicollinearity on the PLS model parameters of a more complicated model structure: The EPSI rating framework. This will be done by means of a simulation study based on empirical observations of correlation structures in the Danish version of the EPSI ratings.
Original languageEnglish
Publication year2009
Number of pages6
Publication statusPublished - 2009
Event6th International Conference on Partial Least Squares and Related Methods - Beijing, China
Duration: 4 Sep 20097 Sep 2009

Conference

Conference6th International Conference on Partial Least Squares and Related Methods
CountryChina
CityBeijing
Period04/09/200907/09/2009

    Research areas

  • PLS path modelling, Multicollinearity, Simulation study, Satisfaction measurement

See relations at Aarhus University Citationformats

ID: 224336