A researcher runs a 3×2 between-subjects two-way ANOVA on exam scores (DV). Factor A = Teaching Method (Lecture, Flipped, Hybrid). Factor B = Study App (No App vs App). SPSS output:
Levene’s Test of Equality of Error Variances: F(5, 174) = 1.31, p = .262
Tests of Between-Subjects Effects (Type III SS):
- Teaching Method: F(2, 174) = 3.22, p = .042, partial η² = .036
- Study App: F(1, 174) = 1.08, p = .300, partial η² = .006
- Teaching Method × Study App: F(2, 174) = 4.18, p = .017, partial η² = .046
Which interpretation and follow-up is MOST appropriate?
Because Teaching Method is significant (p = .042), conclude that teaching method changes exam scores overall and run a Tukey post-hoc test collapsed across Study App; ignore the interaction because it’s smaller than the main effect.
Because the interaction is significant (p = .017), interpret the effect of Teaching Method as depending on Study App usage; follow up with simple effects (e.g., compare teaching methods separately within each Study App condition, with multiplicity control such as Bonferroni/Holm), and report the interaction partial η².
Because Levene’s test is non-significant (p = .262), the main effects are the only results you should interpret; the significant interaction can be ignored as a statistical artifact.
Because Study App is non-significant (p = .300), remove Study App from the model and rerun a one-way ANOVA on Teaching Method; then interpret Teaching Method using post-hoc tests.