Homogeneity of Variance

Question

With reference to the sample data set below (3 treatments ABC, 3 positions 123 and 3 reps), I am interested in knowing if the level of variation in the responses due to treatments A, B and C are the same or sig different. Also if the level or variation in the 3 positions is the same or not. Finally, is there a sig interaction between treatment and position.

            A       B       C
Position 1  1.01    1.03    1.05
            1.02    1.00    1.02
            0.99    1.01    1.01
Position 2  0.98    1.00    1.00
            0.99    1.02    1.00
            1.01    1.04    1.02
Position 3  1.00    0.99    1.01
            1.02    1.02    1.00
            1.01    1.02    1.01

A normal Anova will determine if the means are sig different (or not) but i am interested in assessing the level of variation in the data, is the level of variation in the response affected by treatment and / or position.

Assuming the data is normally distributed I was considering using the Bartlett test to assess the homogeneity of variance for the treatments (3 columns) and then repeating the test for the rows. Is this approach valid? Is there a better way?

Bartlett test is a theoretical textbook test which should not be used in practice. It has no robustness. — kjetil b halvorsen, Oct 14 '16 at 16:11
if not Bartlett what test or methodology would provide a more robust method — user3673417, Oct 14 '16 at 17:52
For homogeneity of variance, maybe the Fligner test (rank-based), or directly some anova test which do not assume constant variance. — kjetil b halvorsen, Oct 14 '16 at 18:49
There is a ton of advice already in here, just search. See http://stats.stackexchange.com/questions/97098/practically-speaking-how-do-people-handle-anova-when-the-data-doesnt-quite-mee/103446#103446 — kjetil b halvorsen, Oct 14 '16 at 18:55

Homogeneity of Variance

0 Answers0