Consider the linear regression model:$$y=\beta_{0}+\beta_{1}x+u$$

The OLS estimators minimze the squared sum of residuals. By the Gauss-Markov theorem, OLS estimators give the lowest variance of estimates in the class of unbiased estimators. However, we also know that estimators such as Ridge estimators exploit the Bias-Variance tradeoff and are able to achieve higher efficiency. Is there an estimator that would maximizes the t-value of a test of the null:$$H_{0}:\beta_{1}=0$$ or the ratio$$\frac{\hat{\beta}}{SE(\hat{\beta})}$$

| cite | improve this question | | | | |
  • 1
    $\begingroup$ In multiple regression? Which t-value are you going to maximize? $\endgroup$ – Glen_b -Reinstate Monica Dec 3 '15 at 16:54

You do know that not all estimators of $\beta$ have a $t$-distribution when divided by an estimator of their sampling variation, right?

If you want to maximize that ratio, just use an estimator that is always equal to some constant, $c$ (pick any value for $c$ that you want). Then, the standard error is zero, and that ratio is infinite.

| cite | improve this answer | | | | |

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.