It is possible that the results concerning interaction effects above arise
spuriously from both an underlying quadratic effect not captured in model 3,
and due to the nature of data transformation technique used. As recommended
by Darlington (1990), therefore, we also tested for curvilinear effects of the
independent variables. We found that the quadratic effects of satisfaction, trust,
and switching barriers on retention were non-significant at the p , 0 :05 level.
We also did the regression analysis using alternative data transformations.
While the results showed different levels of normality and homoscedasticity, the