1
$\begingroup$

If $X_i \sim \mathcal{N}(0,\sigma^2)$ for $i=1,2, \ldots, 2n$ then what is the distribution of $\dfrac{\sum\limits_{i=1}^{n} X_i^2}{\sum\limits_{i=n+1}^{2n} X_i^2}$ ?

Unfortunately I can't figure out how to start working with this problem. Any help?

  • 0
    Do you need to find a density? If $X_i$ are independent, then according to [Wikipedia](http://en.wikipedia.org/wiki/Normal_distribution#Combination_of_two_or_more_independent_random_variables) the distribution is $F_{n,n}$.2011-11-30
  • 0
    In other terms, the ratio R is such that R/(1+R) is Beta(n/2,n/2).2011-11-30

1 Answers 1

0

It's a non-normalized F-distribution.

The F-distribution (on rare occasions called the Beta distribution of the second kind) is usually defined as that of the quotient $$ \frac{\sum_{i=1}^n X_i^2/n}{\sum_{i=1}^m Y_i^2/m} $$ where $n$ and $m$ are called the numbers of degrees of freedom in the numerator and the denominator, and the $X$s and $Y$s are independent and distributed as $\mathcal{N}(0,1)$.

"Non-normalized" means only that the divisions by $n$ and $m$ are not done. It just re-scales the thing.

There is an immense literature on these, and they are introduced in any statistics textbook that discusses ANOVA.