Table 6.

Power (percent) for detecting a treatment effect with a potential period effect based on 4000 simulations in a two-arm, two-treatment crossover study, by the standard deviation of the values after treatment (${\sigma }_{Trt}$), and the period effect (mean μ_P, standard deviation σ_P). Results are for at least 99% of the simulations, except as indicated.

	35/arm, σTrt = 1			50/arm, σTrt = 1.4
Period effect: μP (σ P)	0.0 (0.0)	−0.3 (0.3)	−0.3 (0.1)	0.0 (0.0)	−0.3 (0.3)	−0.3 (0.1)
Nonparametric analyses
Friedman test (NQ only)	93.1	60.7	63.2	72.7	21.6	21.8
Friedman test (NQ and ND)	94.5	68.1	70.3	79.1	33.2	34.2
Mixed models using follow-up data only
Equal variances	98.4	76.2	78.5	81.2	23.1	22.7
Unequal variances	96.6	62.1	65.1	94.4a	54.7a	58.0b
Mixed models using both baseline and follow-up data
Equal variances	84.8	77.5	80.9	56.2	42.4	45.0
Unequal variances	83.5	74.7	78.0	74.1	62.0	65.3

Assumptions: treatment reduces the underlying mean by 0.5 from a baseline mean of 2.5, the standard deviation of the untreated mean is 1, Type I error is 0.05, quantification limit is log₁₀(80) = 1.90.

NQ: non-quantifiable; ND: not detectable.

NQ only : NQ and ND results were combined.

^aResults for 91.5% to 98.5% of all simulations

^bResults for 89.3% of all simulations.