In article <pdoyle-010794095237 at 18.104.22.168> pdoyle at medsun.unige.ch (Patrick Doyle) writes:
>Can anyone tell me when the bonferroni adjustment to samples of 2 groups
>should or should not be carried out.For instance,if I have several areas in
>the brain for which I have taken 6 bilalteral measurements and compare them
>to the same number of measurements and same areas from a separate tested
>group of rats -
>should I first ANOVA then t test with a bonferroni or should I compare the
>weighted totals and then specifically test the areas by the t test -simple
>unpaired .Is it serious to consider the individual areas as being different
>and therefore assuming the various areas as being sampled from an
>heterogeneous tissue?Does this therefore constitute multiple sampling or
>Thanks for any comments.
I am not a statistician! but here goes.
The response to this that suggested using a repeated measures design was good.
I assume however, that you are also comparing multiple areas of the brain?
It is still best to do the whole anova analysis,
If you find significant effects and particularly significant
interactions between brain region and treatment, then you would
be justified in looking at individual brain areas with F tests, which when done correctly, will (like Bonferoni),
correct for multiple comparisons. If you are looking at many different nuclei at once, there could be a problem with sample
size; and you may have to resort to T -tests or in your case individual repeated measures ANovas for each area you look at.
A-priori hypotheses may justify the use of limited non-corrected T-tests?