Non-normality issue

[Ibrahim Al-Jubari]

Hello All,
What do we do if the data is not normally distributed? 
Here is the issue:
Data was collected from university students. Sample size above 200. SEM is intended to be used. We wanted to measure their perception and attitudes towards a given behaviour. When answering to the survey questionnaire, which is a 7 likert scale, they mostly chose 5,6 and 7. In doing preliminary screening, results showed that data is very skewed and the kurtosis is high as well, and thus the normality assumption may be violated. My question: how do we deal with this issue and proceed and data analysis?
Looking forward to seeing your input.
Thank you very much

[Clive Sims]

Hi Ibrahim,

You should find the answer in

Interpreting Quantitative Data With IBM SPSS Statistics (2nd Ed) by…

which I reviewed on Methodspace on May 23 this year. The above is link to Book Reviews.

Hope this helps.

Clive

[Rafael Garcia]

Ibrahim,

The data need not be normally-distributed. The residuals do. If you have many predictors in the model, it should be a non-issue.

There’s a discussion of regression (which SEM is based on) assumptions here: http://andrewgelman.com/2013/08/04/19470/

Best,

Raf

[Ibrahim Al-Jubari]

Clive & Raf,

Thank you so much

[Author]

Posts