Problem Statement:
Given any $n$ events $A_1,A_2, ...,A_n$, prove that the probability of exactly $m$ ($m \le n$) of them happening is
$$P_m - \binom{m+1}{m}P_{m+1} + \binom{m+2}{m}P_{m+2} - \cdots \pm \binom{n}{m}P_n$$
where $P_k = \sum_{1\le i_1<i_2 \cdots <i_k \le n}\Pr(\bigcap_{r=1}^k A_{i_r})$
My thoughts:
This looks quite similar to the inclusion-exclusion principle where we had:
$$\sum_{i=1}^n(-1)^{i+1}P_i$$
but here we have
$$\sum_{i=m}^n(-1)^{i+m}\binom{i}{m}P_i$$
I understand that $P_k$ implies you should sum the probabilities of all the combinations of $A$ of length $k$. Since this already includes the combinations, I don't understand why we need a binomial coefficient before each term.
To find the required probability, one can add all possible probabilities of a combination of $m$ events happening, then subtract those of $m+1$ events happening (since the first one includes this).
The same question has been asked before here and here. But in the first one, the answerer uses the Indicator function (I'm aware of its basic properties, but the answerer does something which is not clear to me), and in the second, the OP themselves provide an answer (but it has a flaw (?) - they take $\Pr(M \cap N^C) = \Pr(M) - \Pr(N)$). So, I'm looking for an answer that at least clearly states how can I proceed with the proof, the reason for the binomial coefficients and uses the Indicator function as less as possible (since till now the author has not discussed it in the book).