SMZ wrote: > What 2-Way ANOVA tool do I use for exploring differences in these > treatments if the variances between groups is clearly not equal > (homogeneous)?
You don't. With cells containing zero you can not trust the interaction (note that your ANOVA only looks at 7 of the 9 cells as the between d.f. is 6; the interaction should have 4 d.f. not 2). One solution is to look at the fully crossed factors that don't have the empty cells (>2 days incubation). A second solution is to run one-way ANOVA and construct contrasts to look at particular Qs of interest. You will still have to be careful about the unequal N. Other solutions may be appropriate depending on the hypotheses you wish to test. For info SPSS would probably either use a harmonic mean (equally weighted means) correction or using method I using regression (which produces similar results to the equally weighted means solution). Thom . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
