$\newcommand{\Cor}{\operatorname{Cor}} \newcommand{\Cov}{\operatorname{Cov}}$My question is why the following expression holds?

$$t = \frac{\hat{\beta_1}}{\operatorname{se}(\beta_1)} = \frac{\Cor(Y,X)\sqrt{n-2}}{\sqrt{1 - \Cor^2(Y,X)}}$$

Here is what I got so far:

\begin{align} \frac{\hat{\beta_1}}{\operatorname{se}(\beta_1)} &= \frac{\frac{\Cov(Y,X)}{S_x^2}}{\operatorname{se}(\beta_1)}\\ &= \frac{\frac{\Cov(Y,X)}{S_x^2}}{\sqrt{\frac{\sum(\hat{Y_i} - Y_i)^2}{n-2}\cdot\frac{1}{\sum(X_i - \overline{X})^2}}}\\ &= \frac{\frac{\Cov(Y,X)}{S_x^2}\sqrt{n-2}}{\sqrt{\frac{\sum(\hat{Y_i} - Y_i)^2}{\sum(X_i - \overline{X})^2}}}\label{a}\tag{1} \end{align}

\begin{align} \frac{\Cor(Y,X)\sqrt{n-2}}{\sqrt{1 - \Cor^2(Y,X)}} &= \frac{\frac{\Cov(Y, X)}{S_xS_y}\sqrt{n-2}}{\sqrt{1 - \left(\frac{\Cov(Y, X)}{S_xS_y}\right)^2}} \label{b}\tag{2} \end{align}

From $\ref{a}$ and $\ref{b}$, we then need to show:

\begin{align} \frac{1}{S_x\sqrt{\frac{\sum(\hat{Y_i} - Y_i)^2}{\sum(X_i - \overline{X})^2}}} &= \frac{1}{S_y\sqrt{1 - \left(\frac{\Cov(Y, X)}{S_xS_y}\right)^2}}\\ &=\frac{1}{S_y\frac{\sqrt{S_x^2S_y^2 -\Cov^2(Y, X)}}{S_xS_y}}\\ &=\frac{1}{\frac{\sqrt{S_x^2S_y^2 -\Cov^2(Y, X)}}{S_x}}\label{c}\tag{3} \end{align}

I tried to expand the $\Cov(Y, X)$ term and left $SSE$ and $SSX$ terms in $\ref{c}$, but there is no any further process.

I am wondering how to continue from $\ref{c}$, or my initial direction of the proof is not correct?


The trick is the following equation:$$\sum(\hat{Y_i} - Y_i)^2 = SSE = (1 - R^2)SST$$

This is why there is a $\sqrt{1 - \Cor^2(Y,X)}$ term in $(2)$. The whole proof is below:

$$ \begin{align} t &= \frac{\hat{\beta_1}}{\operatorname{se}(\beta_1)}\\ &= \frac{\frac{\Cov(Y,X)}{S_x^2}}{\sqrt{\frac{\sum(\hat{Y_i} - Y_i)^2}{n-2}\cdot\frac{1}{\sum(X_i - \overline{X})^2}}}\\ &= \frac{\frac{\Cov(Y,X)}{S_x^2}}{\sqrt{\frac{(1 - R^2)SST}{n-2}\cdot\frac{1}{\sum(X_i - \overline{X})^2}}}\\ &= \frac{\frac{\Cov(Y,X)}{S_x^2}}{\sqrt{\frac{(1 - R^2)S_y^2(n-1)}{n-2}\cdot\frac{1}{S_x^2(n-1)}}}\\ &= \frac{\frac{\Cov(Y,X)}{S_x^2}}{\frac{S_y}{S_x}\frac{\sqrt{1 - R^2}}{\sqrt{n - 2}}}\\ &= \frac{\frac{\Cov(Y,X)}{S_yS_x}\sqrt{n-2}}{\sqrt{1 - R^2}}\\ &= \frac{\Cor(Y,X)\sqrt{n-2}}{\sqrt{1 - \Cor^2(Y,X)}} \end{align} $$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.