Table 2.

Goodness of fit statistics for all LPA models.

Classes	Loglikelihood (df)	AIC	Adjust BIC	Entropy	LMRT	BLRT
2	−1829.96 (10)	3679.93	3686.32	0.791	190.68***	−1929.41****
3	−1781.83 (14)	3591.66	3600.61	0.812	92.30***	−1829.96***
4	−1743.15 (18)	3522.29	3533.80	0.901	74.18***	−1781.83***
5	−1745.32 (22)	3534.63	3548.69	0.830	−4.159	-

*** p < 0.001.

AIC Akaike’s information criterion, Adjust BIC Adjusted Bayesian Information Criterion, LMRT Lo–Mendell Rubin adjusted likelihood ratio test, BLRT, Bootstrap likelihood ratio test.

Lower AIC and BIC indicated better fit. Higher Entropy indicates greater accuracy in allocating members to latent classes. Significant LMRT and/or BLRT indicates improvement over model in which k = k-1. All bootstrapped models were estimated with five draws. Five class model, BLRT was not estimated as the loglikelihood value for the model with one less class was larger than the loglikelihood value for the estimated mode.