Table 2.

Results of simulation study for the DDT deficiency experiment with 15 mice per group and 18 measurements per mouse²⁸.

Scenario	ρ	True ${\beta }_2$	Model	Mean estimated ${\beta }_2$ (IQR)	Coverage	Power	Type I error**
1	0	−0.002	Ind	−0.0020 (−0.0025, −0.0015)	0.9460	0.7307	0.0520
			AR-1	−0.0020 (−0.0026, −0.0015)	0.9490	0.7243	0.0480
			CS	−0.0020 (−0.0025, −0.0015)	0.9360	0.7300	0.0570
			IND-I	−0.0020 (−0.0030, −0.0009)	0.9490	0.2767	0.0520
			Mixed AR-1*	−0.0020 (−0.0026, −0.0014)	0.9567	0.6648	0.0389
2	0	−0.0096	Ind	−0.0096 (−0.0102, −0.0091)	0.9430	1.0000
			AR-1	−0.0096 (−0.0102, −0.0091)	0.9460	1.0000
			CS	−0.0096 (−0.0102, −0.0091)	0.9370	1.0000
			IND-I	−0.0095 (−0.0106, −0.0086)	0.9510	1.0000
			Mixed AR-1*	−0.0096 (−0.0101, −0.0091)	0.9570	1.0000
3	0.5	−0.002	Ind	−0.0020 (−0.0027, −0.0014)	0.8760	0.6830	0.1160
			AR-1	−0.0020 (−0.0027, −0.0014)	0.9450	0.5320	0.0457
			CS	−0.0020 (−0.0027, −0.0013)	0.9090	0.5910	0.0840
			IND-I	−0.0020 (−0.0033, −0.0008)	0.8630	0.3350	0.1340
			Mixed AR-1*	−0.0020 (−0.0027, −0.0013)	0.9523	0.4848	0.0429
4	0.5	−0.0096	Ind	−0.0096 (−0.0103, −0.0089)	0.8743	1.0000
			AR-1	−0.0096 (−0.0103, −0.0089)	0.9510	1.0000
			CS	−0.0096 (−0.0103, −0.0089)	0.9180	1.0000
			IND-I	−0.0096 (−0.0108, −0.0083)	0.8760	1.0000
			Mixed AR-1*	−0.0096 (−0.0103, −0.0089)	0.9550	1.0000
5	0.85	−0.002	Ind	−0.0020 (−0.0031, −0.0009)	0.6437	0.6337	0.3560
			AR-1	−0.0020 (−0.0031, −0.0009)	0.9483	0.2447	0.0523
			CS	−0.0020 (−0.0034, −0.0006)	0.6867	0.5040	0.3070
			IND-I	−0.0020 (−0.0038, −0.0002)	0.6340	0.4887	0.3590
			Mixed AR-1*	−0.0020 (−0.0030, −0.0010)	0.9554	0.2199	0.0458
6§	0.85	−0.0096	Ind	−0.0096 (−0.0108, −0.0084)	0.6240	1.0000
			AR-1	−0.0096 (−0.0107, −0.0085)	0.9413	1.0000
			CS	−0.0096 (−0.0111, −0.0081)	0.6757	0.9990
			IND-I	−0.0096 (−0.0115, −0.0077)	0.6227	0.9947
			Mixed AR-1*	−0.0096 (−0.0107, −0.0085)	0.9488	1.0000

Covariance matrix structures include independence (Ind), autoregressive (AR-1) & compound symmetry (CS). IND-I corresponds to the model with independence covariance structure and a mouse indicator (fixed-effects model). Mixed AR-1 corresponds to the mixed-effects model with random intercept.

*The percentage of datasets for which the model did not converge was 1.3, 2.1, 5.5, 9.6, 8.7, 14.7 for Scenario 1, 2, 3, 4, 5, 6, respectively. For the scenarios for type I error evaluation, the associated percentages were 1.4, 5.2 and 6.8 for ρ of 0, 0.5 and 0.85, respectively.

**Type I error is derived from corresponding scenarios with ${\beta }_2$ = 0.

^§Scenarios in bold face reflect parameter values actually observed in the experiment.