The first Google hit says:

enter image description here

This highly cited answer says:

Partial r is just another way of standardizing the coefficient, along with beta coefficient (standardized regression coefficient)$^1$. So, if the dependent variable is $y$ and the independents are $x_1$ and $x_2$ then

$$\text{Beta:} \quad \beta_{x_1} = \frac{r_{yx_1} - r_{yx_2}r_{x_1x_2} }{1-r_{x_1x_2}^2}$$

$$\text{Partial r:} \quad r_{yx_1.x_2} = \frac{r_{yx_1} - r_{yx_2}r_{x_1x_2} }{\sqrt{ (1-r_{yx_2}^2)(1-r_{x_1x_2}^2) }}$$

Those are not compatible definitions. Does anyone have a source for the formula in the CrossValidated answer?

  • 1
    $\begingroup$ These are defining different things. The two quotations just happen to use the word "standardizing" in different senses, but both are perfectly clear. $\endgroup$
    – whuber
    Commented Apr 22 at 13:47

1 Answer 1


They are compatible in that you can show that the "standardized coefficient" $\hat{\beta^*}$, defined by your first quotation, is indeed identical to the expression $\hat{\beta}_{x_1}$ in your cited answer. This has been proven in my answer as an intermediate step for validating the equivalence between two definitions of $R^2$. Although your second quotation involved the concept of "partial R", I am assuming it is actually irrelevant because your first quotation solely concerns with the "standardized regression coefficient".

Specifically, using the notations in my answer, I showed that for a general multiple linear regression with $p$ independent variables (where $\hat{\gamma}$ denotes the OLS estimate for the original regression parameters): \begin{align*} & \hat{\beta^*} = s_Y^{-1}\Lambda\hat{\gamma}, \tag{1}\label{1} \\ & \hat{\beta^*} = r_{XX}^{-1}r_{YX}, \tag{2}\label{2} \end{align*} where $\Lambda = \operatorname{diag}(s_{X_1}, \ldots, s_{X_p})$, $r_{XX}$ is the order $p$ sample correlation matrix of $(X_1, \ldots, X_p)$, $r_{YX}$ is the $p$-long column vector of sample correlations between $Y$ and $(X_1, \ldots, X_p)$, and $s_{X_i}, s_Y$ are sample standard deviations of $X_i$ and $Y$. In your case, $p = 2$, and what you are trying to confirm is just that the first component of $\eqref{1}$ and the first component of $\eqref{2}$ are the same (which is of course true given that as two vectors, $\eqref{1}$ and $\eqref{2}$ are proven to be the same).


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.