Table 1.

Latent class fit indices for two- through four-class solutions using symptom occurrence rates for 14 symptoms from the Memorial Symptom Assessment Scale

Model	LL	AIC	BIC	VLMR	Entropy
Two-class	−2,051.58	4,161.17	4,265.20	693.53**	0.89
Three-classa	−1,988.86	4,065.72	4,223.56	125.45*	0.91
Four-class	−1,937.50	3,992.99	4,204.64	102.73ns	0.86

Notes:

^*P<0.01;

^**P<0.001.

^aThe three-class solution was selected because its VLMR was significant indicating that three classes fit the data better than two classes, and the VLMR was not significant for the four-class solution, indicating that too many classes had been extracted. In addition, the four-class solution resulted in many estimation warnings, because thresholds had to be fixed at extreme values for seven items due to small class sizes. This result indicates that the results for the four-class solution are unlikely to generalize to other samples. VLMR, Vuong–Lo–Mendell–Rubin likelihood ratio test for the K versus K-1 model.

Abbreviations: AIC, Akaike Information Criterion; BIC, Bayesian Information Criterion; LL, log-likelihood; ns, not significant.