Table 2.

Indices of fit for the latent class analysis.

Model Parameters	Two class solution	Three class solution	Four class solution	Five class solution
LL	−558.276	−540.205	−533.365	−527.758
AIC	1142.551	1120.41	1120.73	1123.517
BIC	1185.299	1186.176	1209.51	1235.318
aBIC	1144.115	1122.816	1123.977	1127.606
LMR-LRT (2* LL diff)	141.262 (p <0.0001)	36.141 (p <0.0001)	13.681 (p = 0.02)	11.213 (p = 0.041)
Boostrap LRT	141.262 (p <0.0001)	36.141 (p <0.0001)	13.320 (p = 0.02)	10.918 (p = 0.045)
Entropy	0.803	0.832	0.907	0.974

The LL provides an indication of how well the model parameters are able to reproduce the data; in general higher values (closer to 0) indicate better fitting models. The AIC, BIC, and aBIC are information criteria used to assess comparative model fit; in general smaller values indicate a better fit, however, in latent class analysis the BIC has been found to underestimate the number of categories according to Nylund et al. [15] and often the optimal solution has one more class than indicated by BIC. The LMR-LRT and Bootstrap LRT test the null hypothesis that the model with one less class (k-1) fits adequately; thus a rejection of the null for a k class solution indicates that the k-1 class is preferred. A significance level of 0.01 was chosen when comparing solutions using the LMR-LRT and Bootstrap LRT. Entropy is an indicator of classification uncertainty and values closer to 1.0 indicate less uncertainty.

aBIC: Sample size Adjusted Bayesian Information Criteria; AIC: Akaike; BIC: Bayesian Information Criteria; LL: Log Likelihood; LMR-LRT: Vuong-Lo-Mendell-Rubin Likelihood Ratio Test; LRT: Likelihood Ratio Test.