Table 3.
Selection of clusters based on variance explained and model-fit.
| Number of variables | Variance explained by each PC | Total variance by # of clusters | AIC value by # of clusters | |||||
|---|---|---|---|---|---|---|---|---|
| PC1 | PC2 | 3 | 4 | 5 | 3 | 4 | 5 | |
| 12 | 57.4% | 13.1% | 53% | 58% | 62% | 2, 683.7 | 2, 442.2 | 2, 149.8 |
| 9 | 68.5% | 16.5% | 64% | 69% | 73% | 1, 426.4 | 1, 114.7 | 879.2 |
| 7 | 75.6% | 14.1% | 70% | 76% | 79% | 734.3 | 452.9 | 228.0 |
| 5 | 76.4% | 17.6% | 68% | 75% | 79% | 475.6 | 229.1 | 44.7 |
Explained variance is presented in %. Values closer to 100% indicate greater variation explained.
AIC, Akaike's Information Criterion. A lower AIC value indicates a better model when the clusters were used as predictor variables in multivariate ANOVAs based on the different outcome variables (of 12, 9, 7, and 5 dimensions).