Keep forgetting…what is multiple R?
When you square it => (multiple R)^2 = R^2 - this is meassure how the independent variables explain the dependant variable.
1 Like
thx pitmaster1. that was helpful.
I think you inversed the relationship.
Multiple R is the correlation between actual and predicted values of the dependant variable.
R2 is the model’s accuracy in explaining the dependant variable.
I think there is a testable difference.
Yes, that sounds right. thx straightjacket.
thx straightjacket for correction.
This meassure is of course also useful for multiple regression.
For exam purposes you need to know one thing : (multiple R)^2 = R^2
so, to conclude:
multiple R = R i.e correlation coefficient
and Multiple R^2 = coefficient of determination ?
‘Multiple R’ is the same ‘r’ (correlation coefficiant) for regressions with 1 independent variable. Also computed as: slope sign SQRT(R^2).
Got it.
What about for regressions with 2+ independent variables? What is its relationship with “r” of the regression? Is there no Multiple R (aka r)? Is there an ‘r’ but no Mutiple R? Is multiple R invalid?
Curriculum specifcally tells us that Multiple R = r = correlation coefficient of regression with 1 independent variable but have not found the text where it digs deeper into its relationship with multiple regression.
Also just to confirm, r is correlation coefficent of entire regression including intercept term (not just independent variable)?
Thanks in advance everyone.
Yes, exactly.
If I’m not mistaken, for a simple regression, multiple R = |r|; i.e., multiple R is not negative.
I believe you are correct. Is multiple R useless for regressions with multiple independent variables? (i.e. Multiple R =/= |r| ) ?
1 Like
In a simple linear regression, R-squared is the square of the correlation between Y and X. You determine the sign of the correlation (if going from R-squared to r) by looking at the slope estimate’s sign (or by a plot of the data).
Multiple R and multiple R-Squared (as the names might have you guess) are really only applicable to the case of a multiple regression.
In a multiple regression, multiple R can be viewed as the correlation between the actual and predicted values of the dependent variable. It can only be between zero and one (since it uses a sum of squares in its calculation, and these cannot be negative).
He is correct in the constraint, but the application isn’t for simple linear regression.
Multiple R only exists for those kinds of regressions…Typically, though, there’s no real value added by looking at multiple R if you properly evaluate the model in other ways.
Wonderful. That is where the word “multiple” comes from 
Yes, that’s it.