The TXA2R rs1131882, P2Y1 rs1371097 and GPIIIa rs2317676 three-loci interactions may increase the risk of carotid stenosis in patients with ischemic stroke

Table 3 Comparison of the best models, prediction accuracies, cross-validation consistencies, and P values for carotid stenosis identified by GMDR

Best model^a	Training balanced accuracy	Testing balanced accuracy	Cross-validation consistency	Sign test (P) †
1	0.476	0.531	7/10	7 (0.427)
1, 2	0.614	0.667	9/10	9 (0.156)
1, 2, 3	0.672	0.686	10/10	9 (0.022) ‡
1, 2, 3, 4	0.497	0.598	8/10	8 (0.421)
1, 2, 3, 4, 5	0.623	0.546	7/10	7 (0.712)
1, 2, 3, 4, 5, 6	0.587	0.623	8/10	7 (0.643)
1, 2, 3, 4, 5, 6, 7	0.511	0.472	9/10	5 (0.745)
1, 2, 3, 4, 5, 6, 7, 8	0.632	0.643	5/10	7 (0.524)
1, 2, 3, 4, 5, 6, 7, 8, 9	0.586	0.597	7/10	8 (0.628)
2, 3, 4, 5, 6, 7, 8, 9, 10	0.617	0.486	8/10	6 (0.586)
1, 2, 3, 4, 5, 6, 7, 8, 9, 10,11	0.523	0.677	6/10	7 (0.467)

^ars1131882, rs2317676, rs1371097, rs194149, rs2267679, rs41708, rs701265, rs1439010, rs16863323, rs9859538, rs11871251 are symbolized as 1–11, respectively
GMDR, generalized multifactor dimensionality reduction
† P adjusted for age, gender, hypertension, diabetes, smoking, hyperlipidemia, homocysteine and hemoglobin A1c
‡best model for combination of different variant was defined as P < 0.05 after adjustment with confounding variables

ISSN: 1471-2377