1

Normal distribution in data is one of the assumptions in multiple regression analysis.

In the situation where the Skewness test, Kurtosis and Q-Q plot showed the evidence of normally distributed data, but at the same time K-S and S-W test showed p<0.05 (non-normally distributed data), how should I interpret this kind of output?

Is it Normal or Non-Normally distributed data?

Thanks in advance.

Hisham
  • 81
  • 1
  • 3
  • 4
    Closely related, though with histogram instead of QQ plot (the upshot is similar - somewhat non-normal data can have graphs that look similar to normal ones): [If my histogram shows a bell-shaped curve, can I say my data is normally distributed?](http://stats.stackexchange.com/q/129417/22228) – Silverfish Jan 27 '15 at 23:59
  • 3
    You probably should ask yourself [whether normality testing actually serves any useful purpose for you](http://stats.stackexchange.com/q/2492/22228) since a distribution close to normality may well suffice for whatever you need normality for. – Silverfish Jan 28 '15 at 00:03
  • 3
    Your data are not actually normal. You know this without testing it (failure to reject does not imply otherwise!), but that's not the question you need to answer. In any case you're most likely to reject in large samples, when normality tends to matter least. – Glen_b Jan 28 '15 at 02:34

0 Answers0