Methodspace - home of the Research Methods community

 

Hi,

 

When assessing the normality based on the skewness and kurtosis reading in SEM, which value do we actually refers to - the kurtosis value, or the critical ratio value.

 

Literature indiates that value of 3.0 for skewness and 10 for kurtosis is acceptable (Kline, 2005).

 

Thanks.

Views: 1280

Replies to This Discussion

The value.  The value is a descriptor of how far from normality it is.  The critical ratio is used to get the p-value, which depends on the sample. So for a larger sample, the same degreed of skew (and hence the same problem) will have a larger critical ratio (which makes it seem like a worse problem.)
Thanks Jeremy.

if you want to know about normality, skew is a VERY poor choice

Kolmogorov - Smirnov, avaialble in most stats packages is much better, or try eyeballing a normal percentile plot, also avialble in eg SPSS.

or just produce a simple box & whiskers diagram. if its non-normal and skewed then the 75% and 25% will not be equal distance from median, also 90% and 10%. if it is high kurtosis, but symmetric then outliers will show up.

 

if the reason for non-normality test is to use t-test or ANOVA, then be aware that unequal variance is a wors problem than skew. use unequal variance form of tests

 

best

 

Diana

RSS

Follow us:

© 2014   Created by SAGE Publications.

Badges  |  Report an Issue  |  Terms of Service