I have $n$ independent and identically distributed random variables $(X_1, \dots, X_n)$. Let $F$ be the CDF of their distribution. I also know the CDF of any sum i.i.d. variables that follow the distribution.

Let $k < l$.

I would like to calculate:

$$\forall t \in \mathbb{R} \qquad \mathbb{P}\left(\sum_{i=1}^lX_i \leq t \text{ and } \sum_{i=k}^nX_i \leq t\right)$$

Since both sums have terms in common, I don't see how to proceed.

If necessary, I could make some additional assumptions on the variables, but I would like to keep the result as general as possible.


Denote for compactness $$S_{1,l} = \sum_{i=1}^lX_i$$ and analogously for any other sum, the first subscript denoting the beginning value of the index, the second, the ending value of the index. Then we have

$$\mathbb{P}\left(\left\{\sum_{i=1}^lX_i \leq t \right\} \land \left\{\sum_{i=k}^nX_i \leq t \right\}\right) = \mathbb{P}\Big(\left\{S_{1,k-1}+S_{k,l}\leq t \right\} \land \left\{S_{k,l}+S_{l+1,n}\leq t \right\}\Big)$$

$$=\mathbb{P}\Big(\left\{S_{1,k-1}\leq t-S_{k,l} \right\} \land \left\{S_{l+1,n}\leq t-S_{k,l} \right\}\Big)$$

Now you have two independent sums being smaller or equal from a (function of a) third independent sum. If $f()$ is the density of a sum, and $F()$ the corresponding cdf, and denoting for further compactness $$ S_{1,k-1} = S_1,\; S_{l+1,n} = S_2,\;S_{k,l} =S_3$$

we have using independence

$$\mathbb{P}\Big(\left\{S_1\leq t-S_3 \right\} \land \left\{S_{2}\leq t-S_{3} \right\}\Big) = \int_{-\infty}^{\infty}\int_{-\infty}^{t-s_3}\int_{-\infty}^{t-s_3}f_1(s_1)f_2(s_2)f_3(s_3)ds_1ds_2ds_3$$

$$=\int_{-\infty}^{\infty}f_3(s_3)\int_{-\infty}^{t-s_3}f_2(s_2)\int_{-\infty}^{t-s_3}f_1(s_1)ds_1ds_2ds_3 $$ $$= \int_{-\infty}^{\infty}f_3(s_3)F_2(t-s_3)F_1(t-s_3)ds_3$$


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.