This Exam P sample reference tests Mixture Distributions. This is a mixture-Poisson variance calculation. The average conditional variance is 5.10 and the variance of the conditional means is 16.09, giving 21.19, choice C.
These notes identify the calculation error associated with each wrong letter when that error is reproducible.
AThe value 5.10 is only E[Var(X|T)], the average within-class Poisson variance. It omits the 16.09 created by different class means.
BThe value 16.09 is Var(E[X|T]), the between-class component alone. It omits the within-class component 5.10.
DThe value 42.10 is E[Lambda squared]. A second raw moment is not a variance; 5.10 squared must be removed before adding the within-class term.
EThe value 47.20 is E[X squared]=E[Lambda+Lambda squared]. It omits subtraction of E[X] squared=26.01.
Original practice · fully worked
Original variant: variance under a random gain mode
An instrument has standardized noise U with E[U]=0 and Var(U)=1. Independently, a high-gain indicator B is Bernoulli with success probability 0.40. The recorded offset is Y=(1+B)U+3B. Calculate Var(Y).
A 1.20
B 2.16
C 2.20
D 3.16
E 4.36
Variant answer in brief
Conditioning on the gain mode gives average conditional variance 2.20 and variance of conditional means 2.16. Their sum is 4.36, choice E.
Setup
Setup
Condition on the binary gain indicator and use independence of U and B.
E[Y∣B]=3B
Var(Y∣B)=(1+B)2
Model
Model
Decompose the variance into within-mode noise and between-mode offsets.
Var(Y)=E[Var(Y∣B)]+Var(E[Y∣B])
Compute
Compute
Average over the two Bernoulli states and add the variance of the conditional offset.
The 3108-page Probability Proof Manual reorganizes 718 verified Exam P solutions by syllabus skill and adds formula proofs, error patterns, and original worked practice.