This study examines the problems that arise when testing the main effects and interaction effects of interval-scaled variables in regression analysis, and proposes solutions. When an interaction term is included in a regression model, the following problems occur in evaluating main effects. First, conclusions can differ considerably depending on linear transformations of the scale. Second, severe multicollinearity arises. Third, the interpretation of main effects becomes ambiguous. As solutions to these problems, subgroup analysis, orthogonal centering, and mean centering are presented and compared. As a result, mean centering was found to be the most useful tool for resolving these problems.