Moderation analysis is frequently applied to the studies of psychology and other social science disciplines. Empirical researchers working on experiments as well as questionnaire surveys are often interested in moderation effect because it can help explain how the direction and strength of the relationship between the independent and dependent variables will change. Given that the categorical variable is frequently encountered in social science researches, how to analyze moderation models with the categorical variable becomes a noteworthy issue. In the present study, we consider two scenarios: one is in a questionnaire survey, known as a cross-sectional or between-participant design; the other is in a longitudinal study at two time-points, or a two-condition experiment with the within-participant design. A procedure is proposed and recommended to analyze the moderation effect when the data is collected from the between-participant design and independent variable or moderator is a categorical variable. The first step is to examine whether the moderation effect is statistically significant by testing R2 change with and without the moderation term. If the moderation effect is not significantly different from zero, stop the moderation analysis. In the second step, the omnibus test is used to examine whether the k-1 simple slopes are zero, where k is the number of the categories. If the omnibus effect is not statistically significant, stop the moderation analysis. In the third step, the pairwise test is used to determine which of the k-1 simple slope is statistically significant. There are two pairwise test methods, namely the pick-a-point approach and Johnson-Neyman (J-N) approach. An example is given to illustrate how to conduct the proposed procedure by using SPSS macro PROCESS software. When the data is collected from the two-condition within-participant design, we may presume that every participant is assigned to both experimental treatments (X), and the dependent variable (Y) is observed under each condition. According to the general data input format (such as in SPSS), there is no X variable, and Y have two columns of values. So the above moderation analysis procedure is not suitable for this design. Then, another procedure is proposed and recommended for such kind of data to analyze the moderation effect, in which the only X is a categorical variable. The first step is to regress the difference in the repeated measured dependent variable Y2-Y1 on moderator Z. If the regression coefficient is not statistically significantly different from zero, stop the moderation analysis. In the second step, a simple slope test is conducted by the pick-a-point approach or Johnson-Neyman (J-N) approach. A second example is given to illustrate how to conduct the proposed procedure by using SPSS macro MEMORE software. Directions for future studies on categorical moderation are discussed at the end of the paper. The above methods and steps could be expended to more complicated moderation models, such as the moderated mediation model with a multi-categorical independent variable or moderator, the additive moderator model, and the multiplicative moderator model.