Table 2.

Percentage of clustering object errors ($OE$) based on 1000 simulation replications when data are generated from (26) with respect to $k$-means (K), convex clustering (Convex), the EM algorithm, and our AIC and BIC selectors in generalized $k$-means.

$\sigma$	$c$	$k$	$n_0=50$				$n_0=100$
			K	Convex	EM	BIC	K	Convex	EM	BIC
0.5	10	2	48.7	46.9	10.0	$\mathbf{0.3}$	47.4	46.9	33.6	$\mathbf{0.0}$
		3	47.7	47.1	14.5	$\mathbf{0.2}$	49.5	49.2	33.1	$\mathbf{0.3}$
	20	2	49.7	49.5	12.8	$\mathbf{1.3}$	49.5	49.4	33.1	$\mathbf{0.3}$
		3	49.8	49.7	12.8	$\mathbf{0.8}$	50.1	50.0	31.8	$\mathbf{0.2}$
1.0	10	2	48.8	48.5	13.1	$\mathbf{0.4}$	49.0	48.6	33.6	$\mathbf{3.1}$
		3	45.6	44.5	14.9	$\mathbf{0.2}$	46.7	45.8	36.4	$\mathbf{0.2}$
	20	2	49.8	49.6	13.3	$\mathbf{0.8}$	49.7	49.6	34.5	$\mathbf{3.8}$
		3	48.4	47.8	14.3	$\mathbf{0.7}$	49.0	48.7	36.5	$\mathbf{0.4}$