Table 2.

Mean Bias in Level 1 Standard Errors Across Varying Simulated Conditions by Analytic Technique (1,000 Replications Each).

Number of clusters	Cluster size	ICC = .05			ICC = .10
		OLS	MLM	CB	OLS	MLM	CB
10	10	−4.2	−2.8	−16.5	−3.4	−2.6	−16.2
	30	0.5	0.7	−12.0	0.2	0.5	−12.9
	50	1.9	2.2	−11.5	1.5	2.0	−12.5
20	10	−0.2	0.3	−6.1	−0.4	−0.3	−6.7
	30	−1.2	−1.2	−7.4	−1.8	−1.2	−8.5
	50	−1.0	−0.8	−7.0	−1.3	−0.8	−7.8
30	10	3.3	2.9	−1.4	2.8	2.2	−2.1
	30	1.2	1.0	−3.7	1.1	1.0	−4.2
	50	2.2	2.3	−1.2	1.4	2.6	−2.5
Number of clusters	Cluster size	ICC = .20			ICC = .30
		OLS	MLM	CB	OLS	MLM	CB
10	10	−2.2	−2.8	−16.3	−1.3	−2.8	−16.8
	30	−0.1	0.6	−14.8	−0.5	0.7	−16.7
	50	1.0	1.9	−14.5	0.6	1.9	−16.5
20	10	−0.5	−0.6	−7.9	−0.5	−0.7	−9.2
	30	−2.5	−1.1	−10.5	−3.0	−1.1	−12.2
	50	−1.8	−0.8	−9.4	−2.2	−0.9	−11.0
30	10	2.0	1.8	−3.7	1.4	1.7	−5.1
	30	0.7	1.0	−5.2	0.4	1.0	−6.2
	50	0.4	2.7	−4.7	−0.4	2.7	−6.6

Note. OLS = ordinary least squares; MLM = multilevel model estimated with restricted maximum likelihood using a Kenward–Roger (Kenward & Roger, 1997) standard error adjustment; CB = cluster bootstrapped standard errors using 1,000 replicates; ICC = intraclass correlation coefficient.