Re: F-test: Varying standard deviations for populations in a 2 way ANOVA

On May 10, 6:10 am, AndreHaupt <aha...@xxxxxxxxx> wrote:
Good day,

I'm doing a 2 way ANOVA (2 x 7) to determine whether a certain row and
column combinations provide significantly better results. (Each
combination being a percentage between 0-1.)

The data in csv format:http://ahaupt.googlepages.com/data.csv(20
observations per cell).

Thus, there are 14 populations. Normality probability plots show that
the populations are normally distributed. I get the following standard
deviation values for each:
0.1448 0.1476 0.0753 0.1408 0.1499 0.1058 0.1410
0.1546 0.1477 0.0684 0.1346 0.1569 0.1112 0.1435

According tohttp://en.wikipedia.org/wiki/F-test, the F-test assumes
similar standard deviations for each population. This is clearly not
the case in my situation, even though the majority are around
0.14-0.15.

How would this affect my ANOVA results? (http://ahaupt.googlepages.com/
anova2.csv)

Your thoughts would be much appreciated.

Regards,
Andre

Try analyzing the logs. The means have much the same pattern, but
the sd's are more homogeneous and less correlated with the means.

.

Relevant Pages