Table 1.

Summary of linear models with the highest support (ΔAICc < 2) explaining the change in RTL. Model-averaged estimates of coefficients, unconditional standard errors, and relative importance (sum of AICc weights) were derived for each parameter included the set of well-supported models. Independent model sets included the proportion of days (arcsine square-root transformed) with either (a) shallow (T_subc < 29°C) or (b) deep (T_subc < 25°C) daily torpor bouts.

	parameters in models with ΔAICc < 2			estimate coefficients		95% CI		relative importance
	parameters in models with ΔAICc < 2			estimate	s.e.	lower	upper	relative importance
(a) model sets including a shallow classification of torpor (i.e. T_subc < 29°C)
body mass change (g)	—	—		0.03	0.01	0.008	0.061^a	1.00
shallow torpor (prop. days)	—			0.39	0.15	0.074	0.709^a	0.65
initial RTL	—			−0.15	0.07	−0.988	−0.040^a	0.65
group (warm)		—		−0.14	0.07	−0.286	0.012	0.35
r²	0.52	0.42
AICc	−16.7	−15.5
Akaike weight	0.63	0.35
(b) model sets including a deep classification of torpor (i.e. T_subc < 25°C)
body mass change (g)	—	—	—	0.03	0.01	0.008	0.049^a	1.00
deep torpor (prop. days)	—	—		0.35	0.14	0.052	0.650^a	0.81
initial RTL	—			−0.37	0.21	−0.801	0.066	0.43
group (warm)			—	·0.14	0.07	·0.286	0.012	0.19
r²	0.52	0.45	0.42
AICc	−17.1	−16.8	−15.5
Akaike weight	0.43	0.38	0.19

^aIndicates CI does not include zero.