Currently I am running normality tests on my data. Whilst doing this, my output produces several pieces of information such as:
- 5% trimmed mean
- Skewness and kurtosis
- Kolmogorov-Smirnov statistic
- Histograms and Q-Plots
My question is which piece of information whould be the most informative to determine normality? Some information (Kolmogorov-Smirnov statistic) would suggest violation of normality whilst others (trimmed mean) indicates normality. I have quite a large data set and have read that even if some tests suggests violation of normality, this is common in larger data sets. I'm rather lost with so many messages! :-s

I'm not an expert so someone please correct me if I'm wrong but I'd look at everything. If K-S suggests data is not normal and trimmed mean suggests it is, I'd be looking at the other information you have, what are skewness and kurtosis like? What about your histograms and Q-plots? Did you get a normality plot with your histogram? How well does it fit? As far as I understand, you need to be able to justify your stats so as long as you have a logic and viable reason, you should be fine. I hope that makes sense... and good luck!

======= Date Modified 16 Aug 2011 09:26:08 =======
============= Edited by a Moderator =============
--spam--

======= Date Modified 16 Aug 2011 09:25:50 =======
============= Edited by a Moderator =============
--spam--

Thank you, Skig :-)

You're welcome. Happy to help!