In simple, the Bonferroni correction factor is a strategy to make the alpha
level more conservative in the light of multiple statistical tests that could
derive a statistical outcome just from the testing itself...a spurious
outcome...not due to any real difference in groups under comparison...it is a
multiple-comparison adjustment when several dependent or independent statistical
tests are being performed all at once, and it works by adjusting downwards the
alpha, to account for the number of comparisons being performed. It raises the
standard of proof or put another way, the evidence via the statistical test or
the test statistic must be large enough to fall within the reduced alpha region
of the curve...the so called 'extreme regions' that are normally set at
0.05...this is made smaller by dividing the alpha by the number of
tests/comparisons done so that you result has to be stronger to achieve
significance as the threshold has been moved so that the 'rejection region'
under the curve is now smaller. Remember, alpha (0.05) is really the chance of
making a Type I error (declaring significance when it does not exist) or saying
there is a difference when there is none...so you are trying to reduce the
chance of Type i error due to the multiple tests that raise the risk of a Type I
error...remember the real aim of these tests is to uncover "is the difference I
am seeing via the statistical test real and is it due to the differences in the
groups under comparison and not due to chance alone...did the different
treatments or interventions differ that much to result in a significant
difference or was this difference due to random error"....the Bonferroni raises
the threshold so to speak, so that you are making it harder to declare
significance....so if doing post hoc analysis looking to see where differences
lie (lets say after doing an ANOVA with a significant F test), the Bonferroni
adjusts the alpha down so that the multiple comparisons you are doing between
the difference treatment groups (likely you would be looking at at least 3 in
ANOVA), do not yield significance between treatment arms just by spurious
means....
Best,
Paul E. Alexander
----- Original Message ----
From: Dr Ebtisam <[log in to unmask]>
To: [log in to unmask]
Sent: Thu, May 19, 2011 4:16:20 AM
Subject: Bonferroni correction
Hi all
I am trying to put the meaning of Bonferroni correction tro an example and i
cant, i know its applicable to multiple testing... could any one answer me
kindly, and if possible tro an example to illustrate its application.
Regards
Dr Ebtisam
|