I was reading the article about the Schuette–Nesbitt formula, which is described as "a generalization of the inclusion–exclusion principle", which has both a combinatorial and probabilistic versions. Another website gave a proof for dependent events (pdf download), and found a third that compares it to Waring's Theorem (pdf)

However, I am still confused. I tried finding a clear worked-out example using discrete probabilities (for simplicity) that the steps are clear from one line to the next - to help in overall understanding of the formula.

Is there a good reference, or an answer that can give a short worked-out example?


I have found an example in following book and my answer is a modified version of the Sec 8.4,8.6 of the book in order to make it concise and clear.

Gerber, Hans U. "Life insurance." Life Insurance Mathematics. Springer Berlin Heidelberg, 1990.

$B_1,\cdots B_n$ are arbitrary events. $N$ is a random variable ranging over $\{0, 1, ... , m\}$. For arbitrary real coefficients $c_1,\cdots c_m$, the Schuette–Nesbitt formula is the following operator identity between shifting operator $E:c_n\mapsto c_{n+1}$ and difference operator $\Delta:c_n\mapsto c_{n+1}-c_{n}$. By definition they are related via $E=id+\Delta$, the SN formula is $$\sum_{n=0}^{m}c_n\cdot Pr(N=n)=\sum_{k=0}^{m}[\Delta^{k}c_0]S_k$$ where $S_k=\sum_{j_1,\cdots j_k}Pr(B_{j1}\cap\cdots \cap B_{jk})$ is the symmetric sum among these $n$ events and $S_0=1$. Note that $[\Delta^{k}c_0]$ means difference operator acting on $c_0$. For example, $[\Delta^{2}c_0]=\Delta^{1}(c_1-c_0)=\Delta^{1}(c_1)-\Delta^{1}(c_0)=(c_2-c_1)-(c_1-c_0)=c_2-2c_1+c_0$. Both operators are linear and hence they have representations in terms of matrix, therefore they can be extended to polynomial rings and modules (since these two objects have "basis",loosely speaking.) $$E=\left(\begin{array}{ccccc} 0 & 0 & 0 & \cdots\\ 1 & 0 & 0 & \cdots\\ 0 & 1 & 0 & \cdots\\ 0 & 0 & 1 & \cdots \end{array}\right) $$ $$\Delta=\left(\begin{array}{ccccc} -1 & 0 & 0 & \cdots\\ 1 & -1 & 0 & \cdots\\ 0 & 1 & -1 & \cdots\\ 0 & 0 & 1 & \cdots \end{array}\right)$$

The proof makes use of the indicator trick and expansion of the operator polynomial $\prod_{j=1}^{m}(1+I_{B_j}\Delta)$ and the fact that $I_A\cdot I_B=I_{A\cap B}$ and $\Delta$ commutes with indicators, I will refer you to Gerber's book.

If we choose $c_0=1$ and all other $c_1=c_2=\cdots=c_n=1$, then SN formula becomes the inclusion-exclusion principle as below: $$\sum_{n=1}^{m} Pr(N=n)=\sum_{k=0}^{m}\Delta^{k}c_0S_k=c_0 S_0+(c_1-c_0)S_1+(c_2-2c_1+c_0)S_2+\cdots =S_1-S_2+S_3+\cdots+(-1)^{n}S_n=[Pr(B_1)+\cdots+Pr(B_n)]-[Pr(B_1\cap B_2)+\cdots+Pr(B_{n-1}\cap B_{n})]+\cdots+(-1)^n\cdot Pr(S_1\cap\cdots \cap S_n)$$

Waring's Theorem gives the probability that exactly $r$ out of the $n$ events $B_1,\cdots B_n$ occur. Thus it can be derived by specifying $c_r=1$ and all other $c$'s=0. The SN formula becomes $$ Pr(N=r)=\sum_{k=0}^{m}[\Delta^{k}c_0]S_k=\sum_{k=r}^{m}[\Delta^{k}c_0]S_k$$ because any term $[\Delta^{k}c_0]=0$ when $k<r$, a change of variable $t=k-r$ will yield Waring's formula.

There is an envelope assignment example in Gerber's book you can have a look into, but my suggestion is to understand it in terms of operator algebra instead of probability.


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.