Contingency theory and moderated regression analysis : the effect of measurement level, measurement error end non-linear relations

Moderated regression analysis is generally accepted as the most appropriate way to asses the viability of contingency models. This paper discusses some problems associated with the application of this technique. If interval data are used main effects may not be interpreted. If ordinal data are used the traditional significance tests may produce er-roneous results. If non-linear relations exist spurious interaction effects may be found. The problems surrounding ternative significance tests, measurement instruments. moderated regression analysis are largely solved by using al-mean centering and carefully developed (i.e., highly reliable)