In PROC HPCLUS, which initialization method assigns initial cluster seeds by selecting observations that are maximally separated from each other?
Select an answer to reveal the explanation.
Short Explanation
You want the initialization that spreads seeds as far apart as possible. INIT=MAXDIST does that by picking observations that maximize the minimum distance between seeds, which often helps clustering converge better.
Full Explanation
The correct choice is INIT=MAXDIST because in PROC HPCLUS it selects initial cluster seeds by choosing observations that maximize the minimum distance between seeds, producing a spread-out initialization that often leads to better convergence. INIT=RANDOM is wrong because it selects seeds randomly from the data. INIT=KMEDIANS is wrong because it initializes using medians, not maximum distance. INIT=FIRSTK is wrong because it uses the first k observations as seeds.