Fig 1.
Univariate simulated data study.
Histograms of randomly selected generated data sets. The solid lines represent the true marginal densities.
Table 1.
Percentage of data sets in which the true number of clusters was found, with the mode of the estimated number of classes in parentheses.
Table 2.
The results of Scenario A2–A4.
Percentage of data sets in which the true number of clusters was found, with the mode of the estimated number of classes in parentheses.
Table 3.
Unequal proportions heterogeneous scenario; a heterogeneous population with three clusters.
λ1 = 0.475, λ2 = 0.475, λ3 = 0.05, μ1 = 1, μ2 = 2, μ3 = 3 and σ1 = σ2 = σ3 = 0.25 (high separation). Percentage of data sets in which the true number of clusters was found, with the mode of the estimated number of classes in parentheses.
Fig 2.
Longitudinal simulated data study.
The left profile belongs to a homogeneous population with one class. The middle one belongs to a population with three classes where classes differ only in intercept, and the right profile belongs to a heterogeneous population with three classes where classes differ both in intercept and slope.
Table 4.
Percentage of data sets in which the true number of clusters was found, with the mode of the estimated number of classes in parentheses.
Table 5.
Percentage of data sets in which the true number of clusters was found, with the mode of the estimated number of classes in parentheses.
Table 6.
Number of latent classes in Hb data for different α and different cut-offs (ψ).
Fig 3.
Hb profiles for four different classes.
Fig 4.
Posterior distribution of non-empty classes (K) for different cut-offs (ψ).