1
$\begingroup$

Suppose I estimate by OLS the following linear model $$ Y_i=\beta_0+\beta_1 X_i+U_i $$ where $Y_i$ denotes the weight of individual $i$ (in pounds), and $X_i$ denotes the height of individual $i$ (in inches).

Let $\hat{\beta}_0$ and $\hat{\beta}_1$ be the OLS estimates. Let $R^2$ be the R-squared from the regression . Let $SER$ be the Root Mean Square Error.

Question: Suppose that instead of measuring weight and height in pounds and inches, the variables are measured in centimeters and kilograms. What are the regression estimates (coefficients, $R^2$, and SER) from this new centimeter-kilogram regression? $$ \begin{aligned} & \text{1 Pound = 0.453 Kilogram}\\ & \text{1 Inch= 2.54 Centimeter}\\ \end{aligned} $$

Thoughts: I found similar questions in this forum (e.g., here and here) but I'm unable to clear up my mind on how to proceed.

(A) With regards to the coefficients, I would suggest to proceed by noticing that $$ \begin{aligned} Y_i=\beta_0+\beta_1X_i+U_i & \Leftrightarrow Y_i*0.453=0.453*\beta_0+0.453*\beta_1X_i+0.453*U_i,\\ & \Leftrightarrow Y_i*0.453=0.453*\beta_0+\frac{0.453}{2.54 }*\beta_1(X_i*2.54)+0.453*U_i \end{aligned} $$ Therefore, $\hat{\beta}^{new}_0=0.453*\hat{\beta}_0$ and $\hat{\beta}^{new}_1=\frac{0.453}{2.54 }*\hat{\beta}_1$

(B) $R^2$ does not change because it has no unit of measure.

(C) Let $\hat{U}_i=Y_i-\hat{\beta}_0-\hat{\beta}_1X_i$ and $SER=\sqrt{\frac{1}{n-2}\sum_{i=1}^n \hat{U}_i^2}$. Note that $$ \hat{U}^{new}_i=Y_i^{new}-\hat{Y}^{new}_i=0.453 *Y_i-\Big[\frac{0.453}{2.54 }*\hat{\beta}_1(X_i*2.54)+0.453\hat{\beta}_0\Big]=0.453* \hat{U}_i $$ Hence, $$ SER^{new}=0.453*SER $$

Are A,B,C correct?

$\endgroup$
1
  • 6
    $\begingroup$ Yes, A, B, and C are correct. The coefficients change according to rescaling, R-squared stays the same and the residuals are rescaled with the DV. $\endgroup$ Aug 26, 2021 at 9:42

0

Your Answer

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

Browse other questions tagged or ask your own question.