Table 4.

Summary of results for Simulation I: The table reports the average attraction rates (percentage of restarts yielding the best-found solution) and average computation times for the LCM, K-means, and K-median methods for each category level of each design feature.

		Average attraction rates			Average computation times
Design feature	Level	LCM	K-means	K-median	LCM	K-means	K-median
Sample size	N = 100	48%	35%	87%	0.28	39.13	0.08
	N = 200	59%	34%	89%	0.70	71.05	0.22
	N = 400	65%	32%	91%	1.69	132.49	0.64
Number of clusters	K = 2	96%	62%	93%	0.08	33.17	0.18
	K = 3	72%	40%	88%	0.44	52.93	0.25
	K = 4	50%	28%	88%	0.83	77.60	0.31
	K = 5	40%	21%	88%	1.27	105.64	0.38
	K = 6	28%	16%	89%	1.81	135.11	0.45
Number of variables	V = 6	47%	11%	85%	1.33	77.34	0.29
	V = 9	57%	37%	91%	0.78	81.34	0.32
	V = 12	67%	53%	90%	0.55	83.99	0.34
Cluster sizes	Equal	67%	41%	89%	0.65	81.14	0.31
	60%	42%	27%	88%	1.29	80.65	0.33
	10%	62%	32%	90%	0.72	80.88	0.31
Error level	5%	51%	35%	79%	0.66	74.53	0.28
	10%	58%	36%	93%	0.86	80.57	0.32
	15%	61%	30%	95%	1.15	87.57	0.35
Overall		57%	34%	89%	0.89	80.89	0.31