Table 1.

Comparisons between the latent class analyses with different number of latent classes.

Model selection criteria	ka=2	k=3	k=4	k=5	k=6
The minimum percentage of 1 class	44.50	22.8	15.46	10.50	7.17
The mean posterior class membership probabilityb	>0.96	>0.91	>0.83	>0.78	>0.77
Entropy	0.84	0.82	0.78	0.77	0.79
Bootstrap likelihood ratio test (k vs k-1)-2 log likelihood (degrees of freedom)	5057.04 (16)c	1010.30 (16)c	274.03 (16)c	182.95 (16)c	141.77 (16)c
Akaike information criteria	27,465.22	26,486.93	26,244.90	26,093.95	25,984.18
Bayesian information criteria	27,639.12	26,750.59	26,598.32	26,537.13	26,517.12

^ak: number of latent classes.

^bThe model with k=2 was selected as the final model considering the highest posterior class membership probability, entropy, statistically significant difference from the model with k=3, and interpretability (ie, more distinctive internet use behaviors between classes).

^cP<.001.