Sufficient Sample Size and Power in Multilevel Ordinal Logistic Regression Models

Table 3

Power of fixed effect estimates of five-category response multilevel ordinal logistic model (method: ML and PQL, category distribution: symmetrical).

Groups	Group size	ICC	ML method			PQL method
			30	5	0.1	0.536	0.551	0.488	0.607	0.611	0.523
0.3	0.561	0.559			0.492	0.596	0.621	0.539
0.5	0.572	0.550			0.501	0.599	0.616	0.531
30	30	0.1	0.624	0.661	0.569	0.661	0.659	0.591
		0.3	0.639	0.639	0.589	0.650	0.672	0.602
		0.5	0.635	0.669	0.594	0.658	0.649	0.611
30	50	0.1	0.715	0.709	0.661	0.739	0.693	0.681
		0.3	0.731	0.721	0.679	0.749	0.716	0.699
		0.5	0.739	0.732	0.703	0.745	0.726	0.693
50	5	0.1	0.850	0.835	0.731	0.886	0.849	0.756
		0.3	0.861	0.849	0.729	0.897	0.863	0.769
		0.5	0.873	0.858	0.740	0.897	0.869	0.763
50	30	0.1	0.871	0.864	0.778	0.903	0.914	0.809
		0.3	0.877	0.856	0.789	0.914	0.895	0.795
		0.5	0.869	0.879	0.768	0.909	0.899	0.799
50	50	0.1	0.891	0.859	0.815	0.956	0.916	0.849
		0.3	0.882	0.880	0.826	0.969	0.931	0.871
		0.5	0.896	0.878	0.831	0.972	0.939	0.878
100	5	0.1	0.992	1.000	0.851	1.000	1.000	0.914
		0.3	0.998	1.000	0.859	1.000	0.997	0.902
		0.5	0.998	0.995	0.867	1.000	1.000	0.909
100	30	0.1	0.995	0.989	0.890	1.000	1.000	0.929
		0.3	0.997	0.998	0.897	1.000	1.000	0.936
		0.5	0.992	0.998	0.879	1.000	1.000	0.939
100	50	0.1	0.989	0.992	0.906	1.000	1.000	0.952
		0.3	0.996	0.996	0.919	1.000	1.000	0.959
		0.5	0.996	0.997	0.899	1.000	1.000	0.951