Hello Jay,

If your concern is with the issue of non-homogeneous group variances (though, in reality, groups of six cases are pretty small for making such judgments), and you're convinced that your data do not conform adequately to the assumption, you could:

1. Use resampling/bootstrap estimates for the effects of interest. (The ideal way would be to run an exact/permutation test, but these are not always available.)

2. Use robust regresssion method for the analogous linear model: Y = b0 + b1X1 + b2X2 + b3X1X2 + e

3. Use a variance-adjusting anova method, such as Welch's method, or the Brown-Forsythe method;

4. If it makes sense to do so, consider a transformation of the DV scores to stabilize observed variances across cells.

Good luck with your work.

Thank you, David.

Is it appropriate to use Two way ANOVA when there are only two groups present. In my case Control and Treatment.

And 2nd question is, isn't Welch's and brown Forsythe method are only for one way ANOVA?

And currently I only have access to SPSS and Graph Pad

Hello again Jay,

1. "Two-way" anova requires two independent variables or factors, each having two or more levels. A model having only one IV such as Control & Treatment could (with appropriate parametric assumptions) be handled via one-way anova or the analogous independent t-test.

2. No, variance-adjusting methods may be used with factorial (e.g., more than one IV/factor) models.

You also have access to the freely available R package, which has more application libraries available for it than any other software framework for data analysis. Have a look at the free package, Jamovi, for a more "point and click" package which is largely (if not exclusively) built on R.

Good luck with your work.

You said you are looking for a difference. I assume in the variability, so the Levene test shows this.

Hello David, Thank you for your suggestion.

I m sorry for the confusion here.

What I mean to say is, I have two groups, Male (Control and Treatment) and Female (Control and Treatment), each with 6 animals per group, so a total of 24 animals. What I mean to ask is, for this experimental design is Two way ANOVA appropriate if the variance test (i.e., Leven's) is significant?

Actually, currently, I do not have the knowledge to use R for statistics.

Hello again Jay,

Yes, now it sounds as if you have two distinct IVs (treatment and sex of animal). I'd recommend you consider using one of the alternative methods outlined (to ordinary, unadjusted two-way anova) listed in my first post to your query.

I don't know what statistical software you're using, but many will offer at least one of the options listed.

Good luck with your work.