If the questionnaire lacks any accepted measuring unit, my strategy is to divide the scores into quantiles. If you have lots of data, then you can use deciles, but even quartiles are useful. This at least gives you an interpretable score in terms of the sample you studied.

One option: You can standardize the scores especially if you have uneven Likert scales for different questions by converting your scores to a 0-100 linear scale using the (Actual score - Lowest possible score) / (Possible score range) * 100) formula.

E.g. Let's say for question 1, you have a 1--5 Likert scale, and the particular score for that question is 3.

Actual score = 3

Lowest possible score = 1

Possible score range = 5 - 1 = 4

(3 - 1)/(4) * 100 = 50.

I hope other chip in.

Regards.

If the questionnaire lacks any accepted measuring unit, my strategy is to divide the scores into quantiles. If you have lots of data, then you can use deciles, but even quartiles are useful. This at least gives you an interpretable score in terms of the sample you studied.

Appreciate your answer Prof.

I'm now in big trouble and I hope that you can give a portion of your time to read my situation.

Actually, I have two datasets obtained from conducting two surveys using questionnaire techniques for physicians and technicians working on hospital information systems, in order to be able to generalize the main factors affecting their behavior to adopt or reject the cloud systems. each questionnaire consist of different sections, each have different constructs based on a number of items. once I collected the data, I faced the problem of analyzing, in which I don't know if I have to use PLS-SEM and generate two models (one for physician and for technicians), OR I have to use the normal regression and correlation of SPSS using the non-parametric tests.

For now I selected the second choice by using SPSS (I hope that you can help me to confirm if this option is relevant or not). I combined the items scores for each construct using the average and then analyse it using the parametric tests (pearson's correlation and multiple linear regression), and here is the problem, is the ordinal data (my collected data) can be analysed using parametric tests after calculating the mean of each construct, or I have to perform non-parametric tests (spearman's correlation and ordinal logistic regression)? In case of non-parametric test, do I have to use SUM/mode/median to combine each constructs responses?

On the other hand, what do you prefer in my case, using SPSS methods mentioned above OR using PLS-SEM and produce two models?

Sorry for my long questions Prof., I hope you can help me.

Thanks in advance and have a great day ahead