Skip to main content

Table 2 Model fit criteria for deciding the number of classes for the full sample (N = 40,986), then for the multigroup analysis with gender added using the known class function, to determine number of classes for men (N = 25,788) and women (N = 15,198). Derived from Mplus

From: A latent class analysis of health risk behaviours in the UK Police Service and their associations with mental health and job strain

Number of classes Entropy AIC BIC SSABIC VLMR-LRT
(p-value)
BVR Range of class probabilities
Full sample
2 0.71 465,395.85 465,680.35 465,575.47 16,062.07 (0.000) 38 0.87 to 0.97
3 0.78 457,361.07 457,792.12 457,633.22 8024.34 (0.000) 25 0.88 to 0.97
4 0.85 456,432.73 457,010.34 456,797.41 957.04 (0.000) 20 0.60 to 0.99
5 0.71 455,663.65 456,387.81 456,120.86 798.66 (0.000) 7 0.60 to 0.86
6 0.71 455,486.82 456,357.54 456,036.56 209.67 (0.000) 5 0.31 to 0.90
7 0.63 455,363.84 456,381.12 456,006.11 156.12 (0.152) 3 0.29 to 0.92
8 0.64 455,198.67 456,362.50 455,933.47 113.96 (0.001) 2 0.43 to 0.82
Multigroup
5 0.77 505,344.14 506,766.60 506,242.23 - 4 -
  1. AIC Akaike Information Criterion, BIC Bayesian Information Criterion SSABIC Sample-size adjusted Bayesian Information Criterion, VLMR-LRT Voug-Lo-Mendell-Rubin adjusted Likelihood Ratio Test, BVR bivariate residuals