Sufficient Sample Size and Power in Multilevel Ordinal Logistic Regression Models

Table 1

Average relative bias of fixed effect estimates obtained as a function of groups, group size, and ICC, collapsing over the category distribution and number of categories.

Groups	Group size	ICC	ML method			PQL method
			30	5	0.1	0.0953	0.0960	0.0523	−0.0925	−0.0858	−0.0810
0.3	0.0790	0.0823			0.0397	−0.1061	−0.0967	−0.0899
0.5	0.0636	0.0675			0.0313	−0.1186	−0.1074	−0.0999
30	30	0.1	0.0708	0.0376	0.0469	−0.0601	−0.0595	−0.0603
		0.3	0.0577	0.0292	0.0496	−0.0677	−0.0714	−0.0797
		0.5	0.0506	0.0263	0.0347	−0.0732	−0.0816	−0.0878
30	50	0.1	0.0314	0.0562	0.0785	−0.0427	−0.0433	−0.0452
		0.3	0.0256	0.0507	0.0778	−0.0539	−0.0551	−0.0720
		0.5	0.0161	0.0466	0.0696	−0.0642	−0.0598	−0.0829
50	5	0.1	0.0750	0.0607	0.0656	−0.0587	−0.0616	−0.0678
		0.3	0.0577	0.0485	0.0558	−0.0645	−0.0669	−0.0763
		0.5	0.0471	0.0489	0.0423	−0.0742	−0.0690	−0.0796
50	30	0.1	0.0175	0.0196	0.0285	−0.0410	−0.0518	−0.0506
		0.3	0.0202	0.0154	0.0219	−0.0474	−0.0443	−0.0559
		0.5	0.0160	0.0154	0.0337	−0.0536	−0.0570	−0.0611
50	50	0.1	0.0281	0.0248	0.0501	−0.0252	−0.0384	−0.0400
		0.3	0.0340	0.0260	0.0333	−0.0391	−0.0370	−0.0447
		0.5	0.0265	0.0243	0.0307	−0.0412	−0.0402	−0.0400
100	5	0.1	0.0319	0.0357	0.0421	−0.0520	−0.0505	−0.0472
		0.3	0.0188	0.0169	0.0255	−0.0549	−0.0416	−0.0357
		0.5	0.0235	0.0188	0.0211	−0.0518	−0.0536	−0.0303
100	30	0.1	−0.0025	0.0066	0.0092	−0.0206	−0.0270	−0.0236
		0.3	−0.0033	0.0044	0.0060	−0.0216	−0.0324	−0.0230
		0.5	−0.0037	0.0019	0.0041	−0.0248	−0.0172	−0.0182
100	50	0.1	0.0064	0.0069	0.0041	−0.0102	−0.0114	−0.0084
		0.3	0.0051	0.0048	0.0025	−0.0105	−0.0125	−0.0243
		0.5	0.0030	0.0031	0.0042	−0.0148	−0.0120	−0.0205