Table 3.

Comparisons of alternative HLM models†

Model	AIC	-2 Log Likelihood	Chi-Square Value	df	Critical Chi-Square (p = 0.05)	Result
No predictors, random intercepts for country and store level only	6,155,857	6,155,849				Base model; intercepts significant; random model appropriate at country level
Add 19 fixed effects at country level	4,323,539	4,323,493	-1,832,356	19	96,439.8	Reject null of no fixed effects
Add 14 random effects at country and store levels	4,283,309	4,283,239	-40,254	14	2875.3	Reject null of no random effects
Add 8 fixed (hypothesized) interaction effects	4,281,425	4,281,331	-1908	8	238.5	Reject null of no hypothesized interaction effects
Add 6 fixed interaction effects for regions and discrete emotions	4,279,246	4,279,136	-2195	6	365.8	Reject null of no interaction effects for regions and discrete emotions

^†Each model is compared to the model above it. The chi-square value is the difference between the -2 Log Likelihood values for the two models. The degrees of freedom are the difference in the number of parameters. The final model has 2,189,063 observations