0
$\begingroup$

Almost all papers I read (in social science) used multiple regression in a "blockwise" manner instead of including all variables at once. I was wondering if it's even possible in our field to not go with the flow and do a multiple regression which includes all variables at once. I know with doing so we won't be able to calculate R-squared of each variable but is it even a problem? Do journals editors care about the exact R-squared of each variable or is it also OK to employ a multiple regression and just have the R-squared of overall model not each variable?

$\endgroup$

2 Answers 2

1
$\begingroup$

I'm not going to try to speak for journal editors, but since you put an SPSS tag on this, I'll note that there's a way in the REGRESSION procedure to get the R2 addition of each predictor on top of all others in a given model. Suppose you have y as the dependent and x1, x2, and x3 as independents. A command like the following will give these results:

REGRESSION /DEPENDENT y /METHOD=TEST (x1) (x2) (x3).

Look for the Subset Tests section of the ANOVA table.

$\endgroup$
1
  • $\begingroup$ Nice to see you on this site, @David Nichols. $\endgroup$
    – rolando2
    Commented Jun 6, 2020 at 19:45
0
$\begingroup$

Another useful statistic is the (squared) semipartial correlation: in SPSS, “/stat zpp”. This tells the unique contribution of each predictor to R-squared.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.