Let $X_1\dots X_n$ be independent random variables with moment generating functions (mgf) $M_i(t) (i = 1 \dots n)$.

The chi-squared random variable with ν degrees of freedom has mgf $M(t) = (1 − 2t)^{ −ν/2}$. When $ν$ is a positive integer, the random variable is the sum of $ν$ independent squared standard normal random variables, i.e.$\sum_{i=1}^{v} Z_i^2$.

(i) Let the $X_i$ be chi-squared random variables with $ν_i$ degrees of freedom respectively. Using the result "$Y = \sum_{i=1}^{n} a_iX_i$ has mgf $M_Y (t) = \prod_{i=1}^nM_i(a_it)$", show that $Y =\sum_{i=1}^n X_i$ is chi-squared with $\sum_{i=1}^n ν_i$ degrees of freedom. Assuming the $ν_i$ are positive integers, give intuition for this result by describing a scenario by which $Y$ might be generated.

HI, would someone please solve for me? It is not my homework, just one of the problems from the past exam. I've tried substituting $M_Y (t) = \prod_{i=1}^n (1 − 2t)^{−ν/2}=(1 − 2t)^{−nν/2}$ and what now?

  • 1
    $\begingroup$ You've done the case where all the $\nu_i$ are the same, but the problem has a different $\nu_i$ for each $X_i$. But otherwise you are 95% of the way there... look at what you get for $M_Y$ and compare it to the MGF of the distribution that you want to arrive at. $\endgroup$
    – angryavian
    Oct 30 '17 at 4:28

Just take the product of the mgfs directly: you have that $M_{X_i}(t) = (1-2t)^{-{\frac{\nu}{2}}}$, so

$$ \begin{align} M_{Y}(t) &= \prod_{i=1}^n(1-2t)^{-{\frac{\nu_i}{2}}} \\ &= (1-2t)^{-{\frac{\sum_{i=1}^n\nu_i}{2}}} \end{align} $$

For the second part, recall that the chi-squared distribution is the sum of squares of normal random variables. What happens when you add these sums together?


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.