Table 1.

General binomial (logistic) models predicting membership in the lasting benefit cluster over the hedonic adaptation cluster from use statistics.

Factor		Model 1	Model 2	Model 3
Fixed effects		Intercept	Intercept	Intercept
		timea	time	time
		freqb	freq	freq
		adhc	adh	adh
		freq×dadh	freq×adh	freq×adh
		—e	freq×time	freq×time
		—	—	adh×time
		—	—	freq×adh×time
Model df		5	6	8
Log likelihood		−877.23	−873.28	−872.92
R-squaredf		[.570,.641]	[.572,.644]	[.573,.645]
Akaike information criterion		1764.5	1758.6	1761.8
Residual deviance		1754.5	1746.6	1745.8
Residual df		1295	1294	1292
Comparison with previous model
	χ2	—	7.9	0.73
	df	—	1	2
	P value	—	0.005	0.7

^aAssessment time point, in days (possible values before mean centering: 7, 14, 21, 28, 35, 37, 42, 72, 97, and 132).

^bFrequency of use (in number of days per week; scaled by total intervention length).

^cAdherence to the specific instructions of an exercise within the past week (coded as 1=true, 0=false).

^dIndicates an interaction effect.

^eNot applicable.

^fThe limits of the R² statistics presented here are Cox & Snell’s pseudo- R² and Nagelkerke’s pseudo- R², respectively.