repeated ANOVA 3, correlation between sd and mean = problem ?



hi all,

here the problem :

I have 16 experimental conditions (3 factors => 4*2*2 lvl) and 14 subjects (n=14). dependent variable = reaction time (RT)
repeated expermiental plan, each subject in all the 16 conditions.

To applied an ANOVA, you have, in this case to control "sphericity". I use Mauchly's test via SPSS and I use an huynh-feldt correction if needed.


But I have a strong corrélation (R²= 0.92) between mean RT and standard deviation.
The more the RTs increase the more the sd increase.

My questions are :

Is this correlation a problem for me accordings to my data and the test i use (3 way within subject anova)

If so, does correction related to sphericity correct the problem ?

Shoul I transform my data to normalise distributions (using for instance, log, square transforme ) and do my anova afterwards ?