You don't need to do a normality test; it's non-normal. A Likert scale can never generate normally distributed data. Nor can it generate continuous data, so it's hard to see how Gunver's suggestion above can apply.

Worse still, Likert scales are generally ordinal, not interval, so it is usually quite hard to justify _any_ statistics designed for interval-scale data. One has to know, for example, that intervals within the Likert scale can reasonably be treated as having consistent meaning.

I'm afraid you should probably consult a nearby statistician who knows your field; there are methods for inference on ordinal data (eg proportional odds logistic regression etc), but they are not often simple.

try this new approach

Article TO DETERMINE SKEWNESS, MEAN AND DEVIATION WITH A NEW APPROAC...

after you enter the data to the statistical program, lets say SPSS, make sure you have all the entries within the possible range {no outliers}, then do the total score for the Likert scale. Next, check the skewness value, it it recommened not to exceed 0.5 

good luck

You don't need to do a normality test; it's non-normal. A Likert scale can never generate normally distributed data. Nor can it generate continuous data, so it's hard to see how Gunver's suggestion above can apply.

Worse still, Likert scales are generally ordinal, not interval, so it is usually quite hard to justify _any_ statistics designed for interval-scale data. One has to know, for example, that intervals within the Likert scale can reasonably be treated as having consistent meaning.

I'm afraid you should probably consult a nearby statistician who knows your field; there are methods for inference on ordinal data (eg proportional odds logistic regression etc), but they are not often simple.

Obviously, the values on likert scales (e.g 5-point ordinal scales) are never normally distributed.

I agree with Mr. Stephen and Mr. Samuel. I am also working on 5-point Likert Scale. I need some literature proving this fact 'Likert scale data can't be normal'.

Please provide some references in support of this fact.

You don't need literature, only some basic logic. If you represent the Likert data by a random variable (RV) - let's call it L -, then this variable will be a discrete random variable, returning the values x1, x2, x3, x4, x5 with probabilities p1, p2, p3, p4, p5, respectively (the values x1...x5 are arbitraty numeric values of your choice, the only contraint is that they are ordered (x1

Hello Muhammad (et al.). Please bear in mind the distinction between a Likert scale and a Likert-type item--see the link below for details. It's not clear (to me) which one you're talking about. HTH.

http://www.john-uebersax.com/stat/likert.htm

Thanks for the link, Bruce. My previous answer was in fact about an item, not about a scale. However, being a sum of of the numerical values from several items doesn't make the points wrong. Some may argue that the scale would thus be a sum of RVs and therefore the distribution of the scale will be a better approximation of the normal distribution. I don't agree here because the items of a scale are designed to be dependent (they must be - otherwise the scale had no meaing; and the central limit theorem applied to sums of independent RVs!).

However, I have one problem with the Uebersax's explanation (and I think this is not his personal point-of-view but rather a common ... misconception):

"Likert was concerned with measuring psychological attitudes, and wished to do this in a "scientific" way. Specifically, he sought a method that would produce attitude measures that could reasonably be interpreted as measurements on a proper metric scale, in the same sense that we consider inches or degrees Celsius true measurement scales." (bold emphasis by me)

A measurement, scientifically, is a comparison to a standard. Ideally, a measurement can be traced back to counting something (e.g. a number of amplitudes of a photon emitted after exiting some atom; a number of atoms) but sometimes just a standard is defined (for instance the kilogram; btw: the word "metric" comes from "meter", see https://en.wikipedia.org/wiki/History_of_the_metre#International_prototype_metre !). As long as such a standard is missing, the assignment of numbers to the values of a Likert item (and thus to the values of a Likert scale) is arbitrary - it is not a measurement.

I strongly believe that with the incontrovertible fact that data gleaned by applying questionnaire designed on Likert scale is commonly considered as ordinal, one does not need to worry about the normality of the data set, but to conduct a nonparametric test using appropriate test statistic. Consequently, i urge you to worry less about ascertainment of normality.

Hi.

Sorry! although this is not related to the asked question, as I don't know how to attract attention to my question, so I'm writing here for help,

https://www.researchgate.net/post/Is_there_any_non-parametric_method_similar_to_Path-Analysis

Thnks a lot