Following an experiment I have 5 datasets consisting of 3 variables and but
10 results. Some of the data is non-normal. If I run Grubbs test I can
detect outliers which if eliminated results in normal data but a reduced
sample size. As an alternative I can Winsorise the data and get normal
results and still have 10 results. My questions are 1) which method is
preferable and 2) if I Winsorise the data should this be applied to every
dataset even those that are normally distributed.
Thankyou.
Paul
_________________________________________________________________
Want to block unwanted pop-ups? Download the free MSN Toolbar now!
http://toolbar.msn.co.uk/
|