With pre- and post-tests you should use an ANCOVA. The non-parametric alternative is the Quade RANCOVA. See the R script in attachment

I suppose you had different subjects for different methodologies. In attahment an example of the Quade RANCOVA. The RANCOVA in this example has more power than the parametric ANCOVA.

You don't have to test normality because likert is. ********** ordinal**********"""""""""""" and thus cannot be normal. That why things like ordinal regression exist. David Booth

In fact Mr Feys is correct if you replace OLS by ordinal regression. For details see for example see the text shown in the attachment available from the z-library Best wishes, David Booth

Likert scale data are primarily discrete in nature and hence it can’t be thought of as to be normally distributed with common perception. Such data do bear its extremities as the values are bounded on the left and on the right as well!

Therefore, it’s advisable that for ordinal data (Likert-scale type individual responses on questionnaire), we use non-parametric tests such as Spearman's rank correlation, Chi-square test for independence, M-W ‘U’ test or any other non-parametric tests.

However, if we transform it into an interval-scale type data using overall Likert scale scores, we may comfortably use parametric tests such as Pearson's ‘r’ correlation or one-sample / two sample t-tests etc. We must remember that, t-test for two independent samples used for normally distributed scale data, and for ordinal or non-normally distributed data we should ideally use non-parametric tests.

Though, it has been commonly found in many places that t-test for Likert item data is being used, but it's neither appropriate nor wise.

Furthermore, we note that a Likert-type scale is composed of a series of five or even more Likert-type items that represent similar questions combined into a single composite variable-score.

Likert scale data can also be analyzed as an interval data using the means and standard deviations to describe the scale. In such cases usually mean is the best measure of central location.

For testing normality, we may use Shapiro-Wilk (S-W) test. Conventionally, If the Significant value of the S-W Test is > 0.05, the data is assumed to be normal. If, however, it is below 0.05, the data-set is said to be suggestively deviate from a normal distribution.

Just go for normal. The means and standard errors will be close.

Likert scale data is not Normal, the main reason being that measures have boundaries (Normal distributions range from -infinity to infinity). This creates skew, when the values approach the boundaries.

https://schmettow.github.io/New_Stats/glm.html#choosing-randomness

The standard solution is ordinal regression. An alternative is Beta regression.

https://schmettow.github.io/New_Stats/glm.html#rating-scales

Just go for non-parametric tests. Then we need not have to necessarily test for the normality of the data. Non-parametric statistics are called distribution-free statistics because they are not constrained by assumptions about the distribution of the population. NP tests can be used even for nominal data (qualitative data like greater or less, etc.) and ordinal data, like ranked data. One can comfortably use The Wilcoxon Signed-Rank test, which is a non-parametric equivalent test of the One sample t-test. SPSS or other standard software package could be put to use for the said purpose.