Interval estimation for proportions

Interval estimation for proportions

To estimate proportions, percentiles or rates we assume we sample from a binomial distribution with size of n with probability p of the event

Examples

X∼Bin(n,p), find 95%CI for p

X=∑Yi,Y1,…,Yn∼Bernoulli(p)≡Bin(1,p)E(X)=np,Var(X)=np(1−p),n iis the sample size For a large sample size (n≥25), we apply CLTy¯=xn=∑yin=p^,Var(y^)=Var(y)n=p(1−p)nZ=y^−pVar(y^)=p^−pp^(1−p^)n∼N(0,1)P(pl≤p≤pu)=1−αP(p^−pup^(1−p^)n≤p^−pp^(1−p^)n≤p^−plp^(1−p^)n)=1−αP(−Z1−α/2≤p^−pp^(1−p^)n≤p^−plp^(1−p^)n)=1−αCI for p is (p^−)X=∑Yi,Y1,…,Yn∼Bernoulli(p)≡Bin(1,p)E(X)=np,Var(X)=np(1−p),n is the sample sizeFor a large sample size (n≥25), we apply the Central Limit Theorem (CLT)Y¯=Xn=∑Yin=p^,Var(Y¯)=Var(Y)n=p(1−p)nZ=p^−pVar(p^)=p^−pp^(1−p^)n∼N(0,1)P(pl≤p≤pu)=1−αP(p^−pup^(1−p^)n≤p^−pp^(1−p^)n≤p^−plp^(1−p^)n)=1−αP(−Z1−α/2≤p^−pp^(1−p^)n≤Z1−α/2)=1−αCI for p is (p^−Z1−α/2p^(1−p^)n,p^+Z1−α/2p^(1−p^)n)P(pl≤p≤pu)=P(p^−pup(1−p)nN≤p^−pp(1−p)nN≤p^−plp(1−p)nN)CI for p:(p^−Z1−α2p^(1−p^)nN,p^+Z1−α2p^(1−p^)nN)