Table 2.

Comparison of alternative HLM models of walking speed

Model	Interceptb	Trials	TXa_level	TX_slope	FUa_level	FU_slope	Reliability	Deviance
A	60.846**						0.954	667.0
B	75.092**	−1.410***					0.986	591.6
C	75.430**	−1.301***	−3.757*				0.986	586.4
D	74.947**	−1.224***		−0.361**			0.987	588.1
E	74.156**	−1.194***	6.800	−0.802*			0.988	579.2
F	70.236***	0.027		−1.273***	−24.769***		0.988	541.8
G	69.023***	0.268		−1.432***		−1.490***	0.991	525.5
H	70.375***			−1.260***		−1.295***	0.992	527.9

*p < 0.05; **p < 0.01; ***p < 0.001

^a TX treatment, FU follow-up

^bIntercept represents the estimated average walking time at time zero, set at the first week of baseline. The variable trials represent each trial of the study, including baseline, treatment, and follow-up trials; statistically, trials represent the change in walking speed over time, not accounting for treatment. TX_slope is set at 0 during baseline and follow-up, and equal to trials during intervention phases; statistically, this variable represents the change in slope during intervention phases. TX_level is a dichotomous variable which represents the change in level of walking speed across intervention phases. FU_slope represents the change in slope during follow-up. FU_level represents the change in level of walking speed across follow-up observations