Table 6. COSINOR analysis performed on all data for each group.

Type of model and parameters subject to a session effect	df	AIC	BIC	loglik	Test	LRT	p
1. Null model (mo): COSINOR Group effect: none	5	-106533.0	-106493.4	53271.49
2. selected model (ms): Extended COSINOR Group effect: phase	6	-106536.6	-106489.1	53274.32	1 vs 2	5.661363	0.017
Parameters of the best model for temperature rise				Value	SD	t-value	p
Mesor				36.98	0.02	1830.09	<0.001
Amplitude				0.68	0.024	14.57	<0.001
Acrophase				17.49	0.37	47.53	<0.001
Δ Acophase patient vs. control				-1.22	0.51	-2.40	0.016

The table displays data from the controls and patients. In order to investigate the potential presence of group effects, we fitted a cosinor function without group effect (i.e., the entire data). Inclusion of a group effect in the mesor, amplitude, and phase showed that a group effect in the phase is supported by a LRT (and AIC; but not BIC). Model m₀ corresponds to the simple cosinor, m_s cooresponds to a cosinor with group effect in the phase:

The parameters m, a, and p correspond to mesor, amplitude, and acrophase respectively. Δ Acrophase re present the shift of the acrophase of the patient group, The estimated acrophase of the control group takes value 17.489 (i.e., 17h29), whereas the acrophase of the patients is significantly advanced by -1.22 (i.e., 1h13).