I will take a middle ground. You should add interaction (moderation) variables to your model. These are of the form b*x1*x2. You should then build your model according to the attached notes. N.B. Prof Davidian is an expert in this subject. The link is:

https://www4.stat.ncsu.edu/~davidian/st732/notes.html

for a general overview see:

https://www.bing.com/search?q=moderation+and+interaction&qs=n&form=QBRE&sp=-1&pq=moderation+and+interaction&sc=2-26&sk=&cvid=C928740A37C54C7AAA48238E3AF34DFC

The analysis used depends on exactly what you want to show. If the categorical variable has 2 levels, you can treat it as if it is continuous using either moderated regression or an SEM program. The following discusses some strategies for this type of longitudinal data.

Article Longitudinal studies in organizational stress research: A re...

As mentioned, SEM might be appropriate for the problem. Please have a look at the following book:

Todd Little - Longitudinal Structural Equation Modeling

I believe that Structural Equatuion Modeling (SEM) can be good for conducting your research. SEM is a multivariate statistical analysis technique that is used to analyze structural relationships. The following packages—Amos, SAS PROC CALIS, R packages sem, lavaan, OpenMx, LISREL, EQS, Mplus, and PLS+ can help users estimate parameters for a model where the structure is well specified.

Mahfuz Judeh Rainer Duesing Paul E. Spector Isam Alkhalifawi thanks every single one of you for providing such insightful advice ! hoping their usefulness will be reflected in my paper . Have a nice day everyone !!

Good answers

Dear @Vasiliki you mentioned that your data is longitudinal in nature and have been collected at three different times viz. T1 T2, and T3 respectively for the same variables. Now the question arises whether you want to check the difference in path coefficients across T1, T2, and T3. If yes then you can use Multi-Group SEM in Smart PLS. Simple SEM won't serve your purpose.