I saw the following claim in some book without a proof and couldn't prove it myself.
ddpP(Bin(n,p)≤d)=−n⋅P(Bin(n−1,p)=d)
So far I got:
ddpP(Bin(n,p)≤d)=ddpd∑i=0(ni)pi(1−p)n−i=−n⋅(1−p)n−1+d∑i=1(ni)[ipi−1(1−p)n−i−pi(n−i)(1−p)n−i−1]
But I am not very good playing with binomial coefficients and don't know how to proceed.
Answer
Consider the derivative of the logarithm:
ddp[logPr[X=x∣p]]=ddp[xlogp+(n−x)log(1−p)]=xp−n−x1−p,
hence ddp[Pr[X=x∣p]]=(nx)px(1−p)n−x(xp−n−x1−p) and ddp[Pr[X≤x∣p]]=x∑k=0(nk)pk(1−p)n−k(kp−n−k1−p)=x∑k=0(nk)kpk−1(1−p)n−k−(nk)(n−k)pk(1−p)n−1−k.
But observe that (nk)(n−k)=n!k!(n−k−1)!=(k+1)n!(k+1)!(n−(k+1))!=(k+1)(nk+1), hence the second term can be written (k+1)(nk+1)p(k+1)−1(1−pn−(k+1)), which is the same as the first term except the index of summation has been shifted by 1. Therefore, the sum is telescoping, leaving ddp[Pr[X≤x∣p]]=0−(nx)(n−x)px(1−p)n−1−x. All that remains is to observe (nx)(n−x)=n!x!(n−x−1)!=n(n−1)!x!(n−1−x)!=n(n−1x), therefore ddp[Pr[X≤x∣p]]=−nPr[X∗=x∣p], where X∗∼Binomial(n−1,p), as claimed.
No comments:
Post a Comment