Given plans to sample 100 elements from a large heterogeneous population, which approach is likely to yield estimates with the most narrow confidence intervals

Select one:

a.Simple random sampling with replacement

b.Stratified random sampling with two strata where their means are greatly different

c.SRS sampling and regression estimation using an auxiliary variable that does not correlate with Y (the estimated parameter)

d.Simple random sampling without replacement

e.Cluster sampling where cluster means vary quite a bit from each other