Original practice · fully worked
Original variant: conditional covariance in a sample without replacement
A container holds four copper tokens, three graphite tokens, and five ceramic tokens. Three tokens are sampled without replacement. Given that exactly one sampled token is graphite, calculate the conditional covariance between the numbers of copper and ceramic tokens sampled.
- A
-0.4938 - B
-0.4321 - C
0 - D
0.4321 - E
0.4938
Answer and calculation
Conditioning leaves two draws from nine non-graphite tokens. The copper count has hypergeometric variance 35/81, and the ceramic count equals two minus the copper count, so their covariance is -35/81=-0.4321, choice B.
Setup
Setup
After fixing one graphite token in the sample, the other two tokens come from the nine non-graphite tokens.
C∼Hypergeometric(N=9,K=4,n=2) Model
Model
Find the variance of the copper count with the without-replacement correction.
Var(C)=2(94)(95)(87)=8135 Compute
Compute
Use the deterministic relation between the two non-graphite counts.
Cov(C,W)=Cov(C,2−C) Cov(C,W)=−Var(C)=−8135=−0.4320988 Answer
Answer
The conditional covariance rounds to -0.4321.
−0.4321(B)