1
$\begingroup$

Let $X$ be a discrete random variable, with positive probability for non-negative integers.

If $q_k = P(X>k)$ and $Q_X(t) = \sum_{k=0}^\infty q_kt^k$, and the probability generating function defined as $G_X(t) = E(t^X)$. Then I have to show that: $$Q_X(t) = (1 - G_X(t))/(1-t). $$

What I got so far is: $$Q_X(t) = \sum_{k=0}^\infty (1 - P(X\leq k))t^k, $$ then assuming $|t|<1$:

$$Q_X(t) = \frac{1}{1-t} - \sum_{k=0}^\infty P(X\leq k)t^k = \frac{1}{1-t} - \sum_{k=0}^\infty t^k \sum_{j=0}^k P(X = j) $$

$$ = \frac{1}{1-t} - \sum_{k=0}^\infty t^k \sum_{j=0}^{k-1} P(X = j) - G_X(t), $$ taking the parts where $j=k$ on the sum to get $G_X(t)$ , but then I can't solve the sum. Is there a viable way to solve that sum? I've also tried to solve without using the complement, but it looked even harder.

$\endgroup$
3
$\begingroup$

Let's say $p_k = P(X=k)$, which means $q_k=\sum_{j=k+1}^\infty p_j$ and so

$Q_X(t)=\sum_{k=0}^\infty (\sum_{j=k+1}^\infty p_j)t^k$ = $\sum_{j=1}^\infty$ $(\sum_{k=0}^{j-1} t^k)p_j $

then for $0\le t<1$, this equals

$\sum_{j=1}^\infty (\frac{1-t^j}{1-t})p_j =\sum_{j=0}^\infty (\frac{1-t^j}{1-t})p_j$

$ = (1 - G_X(t))/(1-t)$

$\endgroup$

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.