1
$\begingroup$

I'm trying to work this question and I need an explaination ( The starting point ). How do I find the MLE for $\alpha$ and MSE for the MLE for $\alpha$ if we have the following: $ X_{1}, ..., X_{n} $ are constants and $u_1, ...u_n$ are iid $ N(0, \sigma^2 )$ and $\sigma^2$ is assumed to be known. Suppose $ Z_1, ..., Z_n$ are random sample that satisfy $$ Z_j = \alpha X_j + u_j $$

Improve this question
$\endgroup$
  • 1
    $\begingroup$ Somebody should probably mention that to denote real numbers by capital letters and random variables by lower cases is the opposite of the conventions of the field. $\endgroup$ – Did Mar 11 '13 at 7:56
  • $\begingroup$ @Did: Yes, I should have -- but we've found before that this particular convention isn't sufficiently engrained in me :-) $\endgroup$ – joriki Mar 11 '13 at 12:45
  • $\begingroup$ @joriki You simply followed the conventions of the text of the question, which is a reasonable choice too. $\endgroup$ – Did Mar 11 '13 at 16:37
1
$\begingroup$

The density for obtaining $Z_j$, given $\alpha$, is that of $u_j=Z_j-\alpha X_j$, so the likelihood is proportional to

$$ \prod_{j=1}^n\exp\left(-\frac1{2\sigma^2}\left(Z_j-\alpha X_j\right)^2\right)=\exp\left(-\frac1{2\sigma^2}\sum_{j=1}^n\left(Z_j-\alpha X_j\right)^2\right)\;. $$

Thus the maximum likelihood is attained at the minimum of the quadratic function of $\alpha$ in the exponent. Since you asked for a starting point, I'll leave it at that for now.

Improve this answer
$\endgroup$
  • $\begingroup$ @ Joriki, I thought $1/sqrt(2\pi)\sigma$ was suppose to be in front of the exponential? or? $\endgroup$ – Eugene Mettle Mar 11 '13 at 12:41
  • $\begingroup$ @Eugene: It says "proportional to", not "equal to". For finding the maximum likelihood, only the dependence on $\alpha$ is relevant; a factor independent of $\alpha$ has no bearing on the value of $\alpha$ that maximizes the expression. $\endgroup$ – joriki Mar 11 '13 at 16:43

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.