Dr AAA,

The following points pertain:

1.      No populations are exactly normal. At least small deviations from normality always exist (MAYBE excepting simulated data from SPSS or SAS etc.).

2.      No matter how close the population is to normality, you can always increase the sample size enough to detect the deviations from normality that exist, with whatever alpha you have selected.

3.      Therefore, the question is not whether the population you’re working with is normal, but whether it’s normal enough for whatever you want to do with it.

4.      A test of the hypothesis of normality is useful only if there’s high power for detecting important deviations from normality, & low power for detecting lesser deviations.

5.      To perform a sensible test you need to establish what is a minimally important deviation from normality, as well as the implications (costs or payoffs) of each combination of a. decision (reject or not reject approximate normality) & b. degree of normality.

6.      If you don’t have time to establish a. & b., I suggest graphing your data in a QQ plot & deciding informally whether the data are sufficiently approximately normal for how you want to use them.

7.      A recent thread in this group discussed these considerations briefly. Someone identified a paper which sought to identify the minimal degree of non-normality that would lead to certain shortfalls in common procedures which assume normality.

--- On Sat, 7/4/09, informatics2005 <informatics2005@...> wrote:

From: informatics2005 <informatics2005@...>
Subject: [Statisticians_group] k-s test of normality
To: Statisticians_group@...
Date: Saturday, July 4, 2009, 9:34 AM

hi
i did k-s test of normality for set of data on SPSS the result was not normally distributed( p= .009)
and repeated it on (medcalc program) data revealed to be normal (p=.228)
i dont know which to trust
and why there s large difference between the to programs for the same data
and is Lillifore correction which done by spss does this significant change?
please if anyone know,plz answer as soon as possible
thank u

dr.aaa