Hi all,

I'm doing a study on the amount of a protein found in a brain region after different manipulations. Exercise is thought to increase it, a diet supplement is thought to increase it, and stress is thought to decrease it. This left us with a 2(exercise, no exercise) x2 (diet, no diet) x2 (stress, no stress) design, and 8 groups in total. We hypothesized that the exercise and diet manipulations would prevent the decrease in the protein that usually results from stress. The dependent variable is a ratio of the target protein divided by a control amount of protein (values ranging from about 0.01 0.06)

The sample sizes were very uneven. For the 8 groups they were: 10, 8, 13, 8, 5, 6, 6, and 10.

When I ran a 3-way Factorial ANOVA on SPSS, Levene's test came out at p = 0.040. From what I've gathered this means the variances of my groups are too different from one another.

Does this mean my ANOVA results are invalid? Are there any other options for analyzing my data?

Thanks,
Andrew