Simulated unequal probability samples were drawn from a known population to compare with‐replacement and without‐re‐placement sampling procedures. Results indicate that the gains in variance from without‐replacement sampling are not sufficient to offset the computational difficulties.
Key words: