Table 1 Summary of the simulation analyses that were conducted, and the variables that were varied in each set of simulations
Analysis | N | k | Effect size | Covariance | Dimensionality reduction | Cluster algorithms |
|---|---|---|---|---|---|---|
(1) What drives cluster separation | 1000 | – 2 (10/90%) | Δ = 0.3–8.1 | 15 features | – None | – K-means |
– 2 (equal) | – None | – MDS | – Ward | |||
– 3 (equal) | – Random | – UMAP | – Cosine | |||
– Mixed | – HDBSCAN | |||||
(2) Statistical power | 10–160 | – 2 (10/90%) | Δ = 1–10 | 2 features | – None | – K-means |
– 2 (equal) | – None | – HDBSCAN | ||||
– 3 (equal) | – C-means | |||||
– 4 (equal) | ||||||
(3) Discrete versus fuzzy clustering | 120 | – 1 | Δ = 1–10 | 2 features | – None | – K-means |
– 2 (equal) | – None | – C-means | ||||
– 3 (equal) | – Mixture model | |||||
– 4 (equal) |