0
$\begingroup$

I am running a regression to predict a variable Y as follows:

$Y=\alpha+\beta\times x+\epsilon$

I am trying to get a distribution of the expected value of Y given standard errors in the model estimate. To do this, I am planning to make random draws of alpha and beta, normally distributed with mean equal to their estimates and standard deviation equal to their standard error, and feed an X value into this model. The output will generate the distribution of Y. However, I don't know how to get the correlation between $\alpha$ and $\beta$, which I can feed into a Cholesky decomposition to create correlated random draws. How do I find this correlation? Also, does this method make sense?

Thanks!

$\endgroup$
  • $\begingroup$ As a suggestion, I think you could also ask this in statistics stack exchange. Perhaps this can increase your odds of getting a satisfactory answer $\endgroup$ – Nate River Jan 31 '16 at 23:49
0
+250
$\begingroup$

In OLS models the covariance matrix of the coefficients is given by $\sigma^2 (X'X)^{-1}$. So, in your case, $$ X = \begin{pmatrix} 1 & x_1 \\ : & :\\1 & x_n \end{pmatrix} $$ hence, $$ (X'X)^{-1} = \frac{1}{n\sum_{i=1}^n(x_i - \bar{x}_n)^2}\begin{pmatrix} \sum_{i=1}^nx^2_i & -\sum_{i=1}^nx_i \\-\sum_{i=1}^nx_i & n \end{pmatrix}, $$ and $\hat{\sigma}^2=\frac{1}{n-2}\sum_{i=1}^n(\hat{y}_i - y_i)^2.$ Hence, to get the estimated correlation you can use the estimated covariance of $\alpha$ and $\beta$ which is given in the secondary diagonal of $\hat{\sigma}^2(X'X)^{-1}$, and the SE which is the square roots of the main diagonal terms respectively.

I'm not sure that your case is exactly the standard one, but I hope it gives some direction to follow.

$\endgroup$

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.