J. Monlezun, Charles, Two-Dimensional Plots for Interpreting Interactions in the Three-Factor Analysis of Variance Model, American statistician , 33(2), 1979, pp. 63-69
In the two-factor factorial fixed effects ANOVA, no interaction between factors may be demonstrated graphically by connecting population cell means to form a set of parallel piecewise linear curves.The three-factor case is considerably more complicated, because both three-factor and two-factor interactions may be present.However, if no three-factor interaction is assumed, or, in addition, if one or more of the two-factor interactions is also assumed to be zero, then the resulting constrained model may be characterized in a manner similar to the two-factor case.Details are presented for three such constrained models.