Independent solution

How to solve this Conditional Probability question

Setup

Setup

Let F denote selection of the ordinary model and L denote selection of the model that always succeeds. The two models have equal prior probability.

Pr(F)=Pr(L)=12\Pr(F)=\Pr(L)=\frac12
Pr(HHF)=14,Pr(HHL)=1\Pr(HH\mid F)=\frac14,\qquad \Pr(HH\mid L)=1

Model

Model

Apply Bayes' theorem to the two-success evidence, then average the next-trial success probability over the posterior model probabilities.

Pr(LHH)=(1)(1/2)(1/4)(1/2)+(1)(1/2)=45\Pr(L\mid HH)=\frac{(1)(1/2)}{(1/4)(1/2)+(1)(1/2)}=\frac45
Pr(FHH)=15\Pr(F\mid HH)=\frac15

Compute

Compute

The next trial succeeds with probability one under L and one-half under F.

Pr(H3HH)=(45)(1)+(15)(12)\Pr(H_3\mid HH)=\left(\frac45\right)(1)+\left(\frac15\right)\left(\frac12\right)
Pr(H3HH)=910=0.90\Pr(H_3\mid HH)=\frac9{10}=0.90

Answer

Answer

The posterior-predictive success probability is 0.90.

910(E)\boxed{\frac9{10}\quad\text{(E)}}