Creation date: 05/12/98
Authored by: Rich Herrington

Question:

It is common whenever ANOVA is performed to assume that the model is additive as this simplifies testing for main effects and interaction effects. If there is a subject by treatment interaction, this implies a non-additive form of the ANOVA model. How do I test for non-additivity in a one-way randomized blocks design?

Answer:

The test is performed using the RELIABILITY procedure with the /STATISTICS=TUKEY subcommand. For three measures we have:

RELIABILITY VARS=MEASURE1 MEASURE2 MEASURE3
/STATISTICS=TUKEY.

BACK | MAIN

Last updated: 01/18/06 by Karl Ho