Table 2.

Bayesian Information Criterion (BIC) and Model Selection

No. of Trajectory Groups	Drinkers
	Light (n=86)			Heavy (n=104)
	BIC	Null Model^a	2ln(B₁₀)^b	BIC	Null Model^a	2ln(B₁₀)^b
1	−683			−2729
2	−661	1	44	−2663	1	132
3	−669	2	−17	−2642	2	42
4	−681	3	−22	−2632	3	20
5	−697	4	−33	−2632	4	0
6	−707	5	−20	−2646	5	−28

The number of trajectory groups in the comparison model.

ln(B₁₀) ≈ 2ΔBIC, ie, twice the difference in BIC between the 2 models, was used to determine the best model to describe quarterly frequency of binge drinking. Using 2ln(B₁₀) ≥6 as a strong evidence for rejecting the null model,⁸⁰ a 4-trajectory group model in heavy drinkers and a 2-trajectory group model in light drinkers best described the data (indicated in bold).