TABLE 5.

Results of the prediction model for valence.

	Model with main effects			Model with interactions
	Coefficient	SE	p-value	coefficient	SE	p-value
Within-person level fixed effects (weekly fluctuations)
Automaticity previous attendance	0.229	0.053	< 0.001	0.210	0.069	0.002
Duration until re-attendance	−0.084	0.080	0.294	−0.085	0.081	0.292
Frequency of attendance before	0.038	0.028	0.169	−0.039	0.177	0.825
Valence end of class	0.028	0.083	0.733	−0.003	0.108	0.978
Increase in valence during class	0.018	0.063	0.778	0.022	0.065	0.740
Valence × frequency	/	/	/	0.011	0.023	0.629
Between-person level fixed effects
Habit strength baseline	0.298	0.090	0.001	0.303	0.091	0.001
Valence end of class	0.623	0.195	0.001	0.639	0.229	0.005
Increase in valence during class	−0.367	0.181	0.042	−0.376	0.189	0.047
Cross-level interaction: Valence x frequency (slope)	/	/	/	−0.009	0.039	0.819
Random effects (variances)
Residual variance automaticity within	4.438	0.408	< 0.001	4.316	0.414	< 0.001
Residual variance automaticity between	1.747	0.454	< 0.001	2.019	0.773	0.009
Slope frequency	/	/	/	0.014	0.034	0.671
Model fit information
LL	−13499.424			−13499.027
AIC	27058.849			27066.053
BIC	27208.446			27235.597

Results from Mplus 8 using full information maximum likelihood estimation for cases with missing values. N = 1,082 observations from 226 persons, results of maximum likelihood estimation with robust standard errors. SE, standard error; LL, Log-likelihood; AIC, Akaike’s information criterion; BIC, Bayes information criterion.