Table II.

Actual Significance Levels of False Covariate Inclusion or Drug Effect Detection Corresponding to a Nominal Level of 0.05, versus the Number of Patients Per Arm

Models	Number of patients per arm	Type I error rate (%)
One-compartment, IV bolus model	5	8.8
	7.5	7
	12.5	6.8
	25	7.2
	37.5	5.9
	50	6.2
	125	4.6
One-compartment, zero-order input model	2	8.87
	6	6.54
	10	6.23
	20	6.03
	24	6.16
	30	5.46
	36	5.72
	40	5.36
	46	5.66
	52	5.71
	60	4.86
	66	5.24
	80	5
Linear disease, slope effect model	10	8.53
	20	5.42
	30	5.76
	44	5.44
	60	5.16
	80	5.02
	120	4.8
Indirect transit compartment model (FPG-HbA1c)	2	6.5
	5	5.4
	10	4.9
	15	4.3
	20	6.01
	25	6.1
	30	5.7
	35	5.8
	40	5.3
HIV viral load, bi-exponential model	25	5.2
	45	4.9
	65	5
	82	4.9
	100	4.9
	115	4.9
Digoxin two-compartment, linear effect model	4	20.44
	8	15.58
	14	13.34
	20	13.09
	30	13.67
	50	12.66