Table 1.

Linear Model of Performance of Hierarchical Clustering and K-Means Clustering Methods Relative to Latent Class Analysis Clustering by Number of Clusters, Number of Observations, and 144 Combinations of Conditional Probability and by Sample Size

Parameter	B	SE	t	p	95% CI
					Lower	Upper
Intercept	0.66	0.05	14.22	0.00	0.57	0.75
Method
Hierarchical Clusteringa	−0.04	0.03	−1.59	0.11	−0.10	0.01
K-Means Clusteringa	−0.02	0.03	−0.67	0.50	−0.07	0.04
Number of Clusters	−0.14	0.01	−10.93	0.00	−0.16	−0.11
Number of Observations	0.00	0.00	−1.34	0.18	−0.01	0.00
Conditional Probability	0.20	0.01	27.08	0.00	0.19	0.22
Hierarchical × Number of Clustersa	0.03	0.01	2.28	0.02	0.00	0.06
K-Means × Number of Clustersa	0.00	0.02	0.13	0.90	−0.03	0.03
Hierarchical × Number of Observationsa	0.00	0.00	−0.48	0.63	−0.01	0.01
K-means × Number of Observationsa	0.00	0.00	−0.48	0.63	−0.01	0.01
Hierarchical × Conditional Probabilitya	−0.02	0.01	−1.98	0.05	−0.04	0.00
K-Means × Conditional Probabilitya	0.00	0.01	0.01	1.00	−0.02	0.02
SD of arcs in transformed accuracy	−0.24	0.38	−0.63	0.53	−1.00	0.51

Notes. Each data point is the mean of 1,000 repetitions with the outcome being arcsine transformed Cramer's V. Parameters for number of observations and conditional probability are adjusted for interpretability; N refers to number of simulations run. Simulations for sample size n = 20 were not included in this analysis because these simulations did not produce interpretable results with two of the methods.

^aComparison method is LCA