. 2020 Mar 17;22(3):e15790. doi: 10.2196/15790

Table 1.

Best-fitting models for the three dependent variables in the post hoc analyses.

Best-fitting model		Dependent variables
		Weight change (kg)^a		Physical activity tracking frequency (days)^b		Average steps (thousands)^c
Coefficients, beta (95% CI)^d
	Continuous vs pre-stop 1		.40 (0.31 to 0.49)		−1.28 (−1.48 to −1.09)		−.18 (−0.29 to −0.07)
	Continuous vs pre-stop 2		.43 (0.34 to 0.51)		−2.25 (−2.44 to −2.05)		−.31 (−0.42 to −0.20)
Tukey-adjusted post hoc comparisons, beta (95% CI)^d
	Continuous vs pre-stop 1		.40 (0.31 to 0.49)		−1.29 (−1.48 to −1.09)		−.18 (−0.29 to −0.07)
	Continuous vs pre-stop 2		.43 (0.34 to 0.52)		−2.25 (−2.44 to −2.05)		−.31 (−0.42 to −0.20)
	Pre-stop 1 vs pre-stop 2		.03 (−0.06 to 0.11)		−.96 (−1.16 to −0.77)		−.13 (−0.24 to −0.02)

^aFixed effects: time period; random effects: distance to goal weight and user ID.

^bFixed effects: time period and distance to goal weight; random effects: user ID.

^cFixed effects: time period and distance to goal weight; random effects: user ID.

^dCoefficients represent mean differences between the three time periods.