Hello Kiera,

VIF and tolerance will both be 1.0 in SPSS when: (a) your IVs are uncorrelated; (b) you have only one IV. Since VIF and tolerance are the inverse of each other (e.g., tolerance = 1/VIF), if one value is 1.0, the other will be as well.

As a general guide, be wary of step-based solutions (regardless of the type chosen) since: (a) there is no guarantee that you will identify the best k-set ensemble of IVs available; (b) many stat packages (SPSS being one) mis-represent the probabilities associated with the tests for entering or removing a variable; and (c) these are particularly opportunistic, and frequently yield a solution that capitalizes on the idiosyncratic characteristics of your data set. Replication is strongly advised!

Good luck with your work!

David Morse thank you very much for your answer. Can I please also ask, do all the variables have to be normally disturbed to be entered into the model or can I add in some non-parametric variables?

Hello Kiera,

It's the residuals (differences between observed and estimated/predicted values on the dependent variable) that must be normally distributed. You can visually check for normality of residuals quite easily in SPSS (select Plots button in the Linear regression dialog box, then check both boxes for standardized residuals displays). If you want a more rigorous check, then Save the residuals to your worksheet, and check them in the Explore subprogram for normality.

Good luck with your work.

Hi David,

That's brilliant. Thank you for your explanation and advice. Really appreciate it!

Thanks,

Kiera

I have 8 computed independent variables and 1 computed dependent variable by using factor analysis. when tested i got VIF=1 for all variables. is it ok?