hi
in a paper which is being prepared for publication there are 15 crosstabs
(tests)
the referees have suggested that a bonferonni corrrection be applied to
the level at which significance is achieved
it is 'normal' to work to 5%
my questions are: is it necessary with a small number of tests to apply a
correction (say signifigance level/number of tests i.e. 5%/15) - the
articles that i have from various medical journals work to a 5% level
if it is common to use a correction, is there a minimum number of tests
that can be performed before a correction is necessary
what we would like to put is "..... less than 20 statistical tests were
carried out ...., a 5% signifigance level was considered adequate""
does anyone have a reference which would justify this statement?
any help greatly appreciated
ann
----------------------
ann e fitzmaurice
medical statistician
dugald baird centre
department of obs and gynae
university of aberdeen
[log in to unmask]
tel 01224 553876 (direct)
fax 01224 404925
|