Nonlinear and Noncompensatory Models in User Information Satisfaction Measurement


This study applies nonlinear and noncompensatory models to examine how users evaluate their satisfaction with their information systems (IS) environment. Several instruments have been developed in the literature to measure user information satisfaction (UIS). These instruments measure user satisfaction by asking respondents to rate their satisfaction/dissatisfaction with a variety of IS attributes; e.g., EDP services, EDP staff, information product, and involvement in IS development. These responses are then combined linearly to develop a surrogate measure for UIS satisfaction. This linear model is derived from Anderson's information theory (Anderson 1981) and based on the assumption that each attribute judgment has a conditionally monotone relationship with the UIS evaluation. However, the literature on attitude formation and decision making suggests that other nonlinear and noncompensatory models are available to decision makers for combining information and are used frequently in attitude formation. In this study, we use two sets of data to examine the linear model and five nonlinear models of decision making to evaluate whether nonlinear models are more effective in predicting a user's overall satisfaction with information systems. First, responses from faculty members at an academic institution were used to test each model. All the non-linear models were more efficient predictors than the linear models. In addition, two nonlinear models - the multiplicative and the scatter models - best represented the data with square multiple correlations of 0.69 and 0.68, as compared to the linear model which had an R2 of 0.61. Second, data from a previous study (Galletta and Lederer 1989) were analyzed to examine whether nonlinear models were more efficient. Data for this study were collected using the short version of the Bailey and Pearson (1983) UIS instrument. Results of the analysis from the full and cross-validation samples show that nonlinear, noncompensatory models performed at par or better than the linear model.


Information Technology and Cybersecurity

Document Type





Linear models, Measurement, Nonlinear models, User satisfaction

Publication Date


Journal Title

Information Systems Research