Yadunath Bajgai, why do you still describe data as non-parametric? Please review some of the answers in another thread you started on Sep 22.

• https://www.researchgate.net/post/An_appropriate_measure_of_variabilty_of_parametric_data

In particular, see the Sep 24 post by Jochen Wilhelm. As he said there:

It is NOT the data that is or is not parametric. It is the statistical analysis of data that is or is nor parametric.

As for regression models that report t-tests or z-tests on the coefficients, the normality assumption is that the sampling distributions of the coefficients are approximately normal. Let's take OLS regression as an example. It is often said that the errors must be normally distributed. Normality of the errors is a sufficient condition, but the necessary condition is approximate normality of the sampling distributions of the model coefficients--see slides 9-10 in the attached PDF. (I said approximate normality, because as George Box noted, nothing in the real world is truly normally distributed.) There is no one sample size that can be deemed large enough for that to happen, because the shape of those sampling distributions depends on both the shape of the raw score population and the sample size.

HTH.

@ Bruce Weaver, Thank you so much for your valuable answer on 'large enough' sample size. OK, it is the statistical procedure not the data itself - thank you again. Best wishes.

@ Bruce Weaver, Thank you for providing the relevant literature. In my previous response I forgot to acknowledge those. Best regards

As of my understanding large enough to indicate above 30. Then there's will not be much difference in the results either you do parametric or non-parametric tests. Before doing the Analysis you can verify the kurtosis value.

@ Anuraj Nayarisseri , Thank you so much for your response. With your input of the sample size of large enough is >30 and that the results would not differ would between the parametric and non-parametric statistical approaches, I learnt a lot. Best regards