Hello Noud,

The standardized regression coefficient estimates ("beta") may be used to compare the relative emphasis given to individual variables in forming the "best" linear composite (given your set of IVs and your sample). Any variable having a significant regression coefficient may be said to be contributing to the explanatory power of the model, however.

Some folks recommend looking at structure coefficients, which are the correlations of individual variables with the linear composite, in addition to looking at the standardized regression coefficients. The reason is that, in the presence of collinearity, a variable with a relatively low standardized regression coefficient may still have a strong structure coefficient value (la.k.a., loading). However, since in MLR the structure coefficient is simply the zero-order Pearson correlation between the variable and the DV multiplied by a constant (the multiple R), it doesn't matter whether you look at structure coefficients or ordinary Pearson r values.

All of which is to say, both statistics can be helpful in interpreting what's going on with your set of predictors/IVs.

Good luck with your work.

Since you are dealing with Likert based DVs I would suggest that you look at:

https://www.researchgate.net/post/Can_you_help_with_statistical_analysis_advice

You are dealing NOT with a Continuous DV but An Ordinal based DV so ordinal logistic regression NOT OLS based regression applies, so you are using an incorrect technique. If you see my answers to the above questions and the notes I attached you should be fine. If you have further questions, please ask. Best, David Booth PS I am assuming your model is intended to be predictive so selecting best predictors may be important to you. DO NOT use stepwise regression but rather adaptive lasso ordinal regression which is not available in SPSS but IS optimal and is available in R. For a discussion of this in the binary DV and Cox regression cases see the attached. For more information see these links as well:

https://www.google.com/search?q=Adaptive+lasso+for+ordinal+logistic+regression+R&rlz=1C1CHBF_enUS874US874&oq=Adaptive+lasso+for+ordinal+logistic+regression+R&aqs=chrome..69i57.32758j0j1&sourceid=chrome&ie=UTF-8

https://towardsdatascience.com/stopping-stepwise-why-stepwise-selection-is-bad-and-what-you-should-use-instead-90818b3f52df

Since this can be a bit complicated, Please ask if you have questions

As a complement to David and David: are you sure that you work with composites? Are these all one-indicator measurements or have you calculated scores? Isn't it maybe even a factor model?