Table 3.

Comparison of the FDR of the proposed P-value based method and the LRT method with 1, 5 and 10 anchor items, respectively, at the FDR control of 5%. The values are averaged out of 100 replications.

			Small DIF			Large DIF
			High	Medium	Low	High	Medium	Low
$N=500$	Small $d_j$	proposed	0.0167	0.0255	0.0298	0.0192	0.0213	0.0319
		LRT 1	0.0089	0.0071	0.0137	0.0119	0.0148	0.0233
		LRT 5	0.0071	0.0181	0.0267	0.0122	0.0195	0.0394
		LRT 10	0.0033	0.0148	0.0283	0.0027	0.0154	0.0329
	Large $d_j$	proposed	0.0240	0.0222	0.0323	0.0231	0.0249	0.0404
		LRT 1	0.0164	0.0212	0.0267	0.0152	0.0216	0.0280
		LRT 5	0.0124	0.0221	0.0308	0.0128	0.0215	0.0246
		LRT 10	0.0031	0.0219	0.0237	0.0029	0.0159	0.0408
$N=1000$	Small $d_j$	proposed	0.0238	0.0277	0.0349	0.0229	0.0269	0.0425
		LRT 1	0.0087	0.0083	0.0152	0.0083	0.0131	0.0170
		LRT 5	0.0100	0.0217	0.0327	0.0087	0.0218	0.0341
		LRT 10	0.0021	0.0191	0.0389	0.0020	0.0162	0.0408
	Large $d_j$	proposed	0.0217	0.0302	0.0390	0.0227	0.0333	0.0444
		LRT 1	0.0165	0.0166	0.0248	0.0172	0.0193	0.0237
		LRT 5	0.0114	0.0155	0.0249	0.0100	0.0162	0.0250
		LRT 10	0.0007	0.0062	0.0218	0.0013	0.0079	0.0260