I am developing a multiple regression model using three predictor and one outcome variable. All variables use the same scale (0-10). Therefore it should suffice to look at the unstandardized coeffients for interpretation, shouldn't it? Are the coeffecients comparable as to "size" of their effect?
Kind regards
It's always fine to look at unstandardized coefficients. I don't typically use standardized coefficients for much. Comparing the values of different coefficients to each other and interpreting that as a measure of 'importance' (which is what I think you're getting at) is a dicey affair. To do so, you need to assume that the units of the different variables are commensurate (if X1 is meters and X2 is seconds, how many meters equals 1 second?), which is usually not reasonable. The fact that your variables are all scaled 0-10 makes this look less suspicious, but doesn't necessarily make it better in actuality.
A different issue is whether your outcome variable is also 0-10. It sounds like it is. If so, standard (i.e., OLS) multiple regression may not be valid. If the Y data don't approach the bounds and the residuals look normal enough, you should be fine, but otherwise you may want to do ordinal logistic regression.
