Subscribe Bookmark RSS Feed

unique variance for each variable in multiple regression

xiaoyaj0

Community Trekker

Joined:

Apr 25, 2016

In the attached dataset and screenshot, I am trying to find the effect of prior academic achievement and SES on current achievement for students. How could I find the unique variance/semipartial correlation for each variance? I know it is not the square of standard beta for the variable. Any ideas? Thank you!


1 ACCEPTED SOLUTION

Accepted Solutions
jiancao

Staff

Joined:

Jul 7, 2014

Solution

JMP doesn't report semipartial correlations directly, but they can be calculated from regressions.

For example,  to get the semipartial correlation for socioeconomic status, first, run two regressions as shown in the screenshots.

12954_pastedImage_0.png12955_pastedImage_1.png

Then take the sqrt of the difference in RSquare values between the two models, i..e. sqrt(0.723934-0.715378)=0.0925. This indicates the "unique" contribution of socioeconomic status as a predictor.

1 REPLY
jiancao

Staff

Joined:

Jul 7, 2014

Solution

JMP doesn't report semipartial correlations directly, but they can be calculated from regressions.

For example,  to get the semipartial correlation for socioeconomic status, first, run two regressions as shown in the screenshots.

12954_pastedImage_0.png12955_pastedImage_1.png

Then take the sqrt of the difference in RSquare values between the two models, i..e. sqrt(0.723934-0.715378)=0.0925. This indicates the "unique" contribution of socioeconomic status as a predictor.