2
$\begingroup$

$X$ and $Y$ have bivariate normal distribution and have joint pdf
\begin{equation*} f\left( x,y\right) =a\exp \left( \frac{-1}{2}\omega \right) ,\text{where }% \omega =6x^{2}+12y^{2}-16xy-8x+24 \end{equation*} then what are the means of $X$ and $Y$?

I have \begin{equation*} \frac{1}{\left( 1-\rho ^{2}\right) \sigma _{X}^{2}}=6,~\frac{1}{\left( 1-\rho ^{2}\right) \sigma _{Y}^{2}}=12~\text{and }\frac{-2\rho }{\left( 1-\rho ^{2}\right) \sigma _{X}\sigma _{Y}}=-16 \end{equation*} so i can find $\rho ,\sigma _{X}$ and $\sigma _{Y}$ using these equations. But how can i get $\mu _{X}$ and $\mu _{Y}$?

this question asked in an exam and requested to be replied in 2 minutes. Is there a shortcut for this?

$\endgroup$
  • 1
    $\begingroup$ Match the theoretical expression of a joint density of two correlated normals, to what you have, and see where it gets you. $\endgroup$ – Alecos Papadopoulos Apr 15 '14 at 14:09
  • $\begingroup$ i have already tried that way and i know i can get the result using matching. But this question asked in an exam and requested to be replied in 2 minutes. $\endgroup$ – mert Apr 15 '14 at 14:16
  • $\begingroup$ Do the means coincide with the point where the density is maximized? If so you have a shortcut $\endgroup$ – Adrian Apr 15 '14 at 16:18
  • 1
    $\begingroup$ Just minimize the function $\omega$ over (x, y), which I think gives you (6, 4) $\endgroup$ – Adrian Apr 15 '14 at 16:19
  • 1
    $\begingroup$ I think this works: en.wikipedia.org/wiki/Multivariate_normal_distribution the mode of a multivariate normal is $\mu$ $\endgroup$ – Adrian Apr 15 '14 at 16:25
1
$\begingroup$

The means coincide with the point where the density is maximized. See en.wikipedia.org/wiki/Multivariate_normal_distribution: the mode of a multivariate normal is $\mu$.

The shortcut is therefore to minimize the function ω over $(x, y)$, which gives $\mu = (6, 4)$.

$\endgroup$
2
$\begingroup$

Hint: As Adrian points out, we need to minimize $\omega$ as a function of $x$ and $y$. Now, \begin{align} \omega(x) &= 6x^2 - x(16y+8) + \cdots\\ \omega(y) &= 12y^2 - y(16x) + \cdots \end{align} have minima at $$x = \frac{16y+8}{12} = \frac 43y + \frac 23, \quad y = \frac{16x}{24} = \frac{2}{3}x$$ respectively, giving $(x,y) = (6,4)$ as the solution. Here, instead of the calculus-based method of finding minima, I used the fact that $ax^2+bx+c$ has an extremum at the average value of its roots $\frac{-b\pm \sqrt{b^2-4ac}}{2a}$ which average equals $\frac{-b}{2a}$ (no need to fuss with evaluating the square root). I leave it up to you to determine if this is doable in 2 minutes or not.

$\endgroup$

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.