Hi, I have a study with a two-way between groups ANOVA (full factorial design). The interaction effect between A*B is significant. Each factor has two levels (A1,A2, B2, B2). How to conduct a planned contrast to compare A1 (B1vs.B2), and A2 (B1vs.B2)? I know it is possible with syntax, but could not find more information on how to do it with a 2x2 design. Thanks!
Some things in SPSS cannot be done via the menu system. This is one of them. You need to use the COMPARE option an an EMMEANS sub-command that is giving you the estimated marginal means for the A*B cells. Something like this ought to do it:
UNIANOVA Y BY A B
/EMMEANS = TABLES(A)
/EMMEANS = TABLES(B)
/EMMEANS = TABLES(A*B) COMPARE(A)
/EMMEANS = TABLES(A*B) COMPARE(B)
/DESIGN A B A*B.
I've used Y to represent the dependent variable and A & B to represent the two between-Ss factors. The EMMEANS line with COMPARE(A) will generate contrasts of A1 vs A2 within each level of B; and the one with COMPARE (B) will generate contrasts of B1 vs B2 within each level of A.
HTH.
Sorry, forgot to mention, I'm using SPSS. The number of observations in each of the four treatment groups is not exactly the same (48,45,51,51). The simple effects of the interaction were insignificant, but I suppose that's because the 2-way interaction is a crossover. So I'm wondering whether another follow-up analysis such as contrasts makes sense in a 2x2 design (?) as I find it hard to interpret the interaction right now. Thanks a lot!
Some things in SPSS cannot be done via the menu system. This is one of them. You need to use the COMPARE option an an EMMEANS sub-command that is giving you the estimated marginal means for the A*B cells. Something like this ought to do it:
UNIANOVA Y BY A B
/EMMEANS = TABLES(A)
/EMMEANS = TABLES(B)
/EMMEANS = TABLES(A*B) COMPARE(A)
/EMMEANS = TABLES(A*B) COMPARE(B)
/DESIGN A B A*B.
I've used Y to represent the dependent variable and A & B to represent the two between-Ss factors. The EMMEANS line with COMPARE(A) will generate contrasts of A1 vs A2 within each level of B; and the one with COMPARE (B) will generate contrasts of B1 vs B2 within each level of A.
HTH.
Thanks Bruce, that's exactly what I was searching for! Is this then called a planned contrast? (just to correctly refer to it in my paper :))
Hello Lisa. Yes, you can call them planned contrasts (assuming they really were planned in advance). Normally, you would want only one of those EMMEANS lines for A*B, by the way. I showed both, because I wasn't sure which one you wanted. In language folks in Psychology might use, you want the pairwise contrasts for the focal variable at each level of the moderator variable. HTH.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Modeling
- Analyses of Variance