Partial R squares

evm

New Member
#1
Is there a way to generate partial r-squares (automatically) in multiple regression? Assume a Model Y=X1 X2 X3. GLM produces estimates of R square for the whole model - but what if I want the contributions of each X to the model. And I could not use REG, since some independent variables are categorical variables.

Thanks.
 

Dr.D

New Member
#2
Is there a way to generate partial r-squares (automatically) in multiple regression? Assume a Model Y=X1 X2 X3. GLM produces estimates of R square for the whole model - but what if I want the contributions of each X to the model. And I could not use REG, since some independent variables are categorical variables.

Thanks.

If you want to find the amount of variance for each predictor, look at the semi-partial coefficients (you have to request them SPSS under 'Statistics' tab, choose 'Part and Partial correlations'). The coefficients table in the column that is marked 'Part' shows the semi-partial correlations for each predictor. When you square those values for each predictor, you find the amount of unique variance explained by each X/predictor.
 

Dr.D

New Member
#4
Thanks.

Do you know how to do with SAS? I don't have SPSS.
I believe in SAS when you go to perform the regression, you choose the statistics tab below, and you should see semi-squared partial correlations (they will be already squared). Select that and run the regression.

Interpret the values as percentages (.34 = 34%)