Table 6.
Summary of results for Simulation II: Mean cluster recovery precision for each method and each level of each design feature: Part 1 – the average percentage of datasets for which the correct number of clusters was selected.
| Design feature | Level | LCM (AIC2) | LCM (AIC3) | LCM (BIC) | K-means (CH) | K-median (MRPC) |
|---|---|---|---|---|---|---|
| Sample size | N = 100 | 72% | 67% | 52% | 46% | 69% |
| N = 200 | 75% | 81% | 68% | 49% | 77% | |
| N = 400 | 75% | 91% | 82% | 51% | 82% | |
| Number of clusters | K = 2 | 86% | 99% | 100% | 92% | 98% |
| K = 3 | 82% | 94% | 86% | 44% | 87% | |
| K = 4 | 79% | 83% | 68% | 33% | 72% | |
| K = 5 | 66% | 67% | 48% | 39% | 64% | |
| K = 6 | 56% | 54% | 36% | 37% | 59% | |
| Number of variables | V = 6 | 75% | 67% | 50% | 41% | 59% |
| V = 9 | 71% | 80% | 66% | 44% | 80% | |
| V = 12 | 75% | 91% | 87% | 61% | 89% | |
| Cluster sizes | Equal | 81% | 89% | 78% | 62% | 92% |
| 60% | 61% | 65% | 53% | 34% | 55% | |
| 10% | 79% | 85% | 72% | 51% | 81% | |
| Error level | 5% | 84% | 86% | 80% | 84% | 92% |
| 10% | 76% | 82% | 70% | 41% | 77% | |
| 15% | 62% | 70% | 53% | 22% | 59% | |
|
| ||||||
| Overall | 74% | 80% | 68% | 49% | 76% | |