Table 2. Determining the number of latent subgroups using the unconditional LGM.

Latent Class	Log- likelihood	AIC	BIC	SABIC	ACP	LMR-LRT
						2LL	P-value
Male
1	−23 170.1	46 352.3	46 395.7	46 373.6	—	—	—
2	−22 018.4	44 062.9	44 150.4	44 109.1	0.89	46 723.3	<0.001
3	−21 890.3	43 820.5	43 955.3	43 891.7	0.77	42 497.1	<0.001
4a	−21 842.0	43 738.1	43 920.0	43 834.2	0.75	42 093.8	<0.001
(4)b	(−21 239.5)	(42 587.0)	(42 949.7)	(42 778.1)	(0.76)	(41 383.0)	(<0.001)
5	−21 822.1	43 712.3	43 941.3	43 833.3	0.70	41 945.4	<0.001
6	−21 811.9	43 705.8	43 982.1	43 851.8	0.65	41 888.1	<0.001
Female
1	−21 989.5	43 991.1	44 031.3	44 012.3	—	—	—
2	−20 860.9	41 747.7	41 834.9	41 793.6	0.90	46 708.7	<0.001
3	−20 785.4	41 610.8	41 745.0	41 681.7	0.80	42 350.4	<0.001
4a	−20 723.6	41 501.1	41 682.3	41 596.5	0.75	42 170.9	<0.001
(4)b	(−20 144.8)	(40 397.6)	(40 759.1)	(40 587.5)	(0.72)	(41 351.9)	(<0.001)
5	−20 709.6	41 487.3	41 715.5	41 607.4	0.66	41 951.4	<0.001
6	−20 704.8	41 491.7	41 766.8	41 636.5	0.66	41 905.3	<0.001

ACP, average classification probability; AIC, the Akaike Information Criterion; BIC, the Bayesian Information Criterion; LMR-LRT, Lo-Mendell-Rubin adjusted likelihood ratio test; SABIC, sample size adjusted BIC; 2LL, 2 times the log-likelihood difference between k and k − 1 class models.

^aThe model with an optimal number of latent subgroups from the unconditional LGM.

^bThe model-data fit indices from the conditional LGM with covariates adjustment and distal outcome.

All estimates are weighted by a three-stage cluster sampling design of 2013 national YRBS.