Table 3.

Significance of the genomic predictors of parity derived in the discovery phase and evaluated in the validation phase using logistic regression models: estimated coefficients, standard errors, and P values of parity scores and other clinical covariates in the validation phase

	Classification tree	Tree AdaBoost	Support vector machine	Random forest	Neareast shrunken centroids
	β (SE) P	β (SE) P	β (SE) P	β (SE) P	β (SE) P
Parity score^a	1.799 (0.98) 0.0657	4.726 (1.77) 0.0075	3.611 (1.77) 0.0414	4.122 (1.72) 0.0164	4.289 (1.30) 0.0010
BMI	–0.094 (0.13) 0.4640	–0.169 (0.14) 0.2391	–0.155 (0.14) 0.2705	–0.176 (0.14) 0.2242	–0.209 (0.14) 0.1358
Breast density	0.931 (0.72) 0.1931	1.363 (0.81) 0.0923	1.182 (0.76) 0.1201	1.307 (0.79) 0.0959	1.071 (0.80) 0.1816
HRT duration	0.037 (0.07) 0.5914	0.037 (0.07) 0.6063	0.022 (0.07) 0.7496	0.022 (0.07) 0.7591	0.048 (0.08) 0.5246
OC use	1.003 (0.67) 0.1342	0.791 (0.70) 0.2602	0.797 (0.67) 0.2308	0.753 (0.68) 0.2662	0.769 (0.78) 0.3247
Smoking duration	–0.013 (0.02) 0.5331	–0.019 (0.02) 0.3707	–0.013 (0.02) 0.5307	–0.015 (0.02) 0.4637	–0.009 (0.02) 0.7087

^aParity score is a woman's probability of being parous estimated using the genomic predictors of parity derived in the discovery phase. The significance of the parity scores was evaluated in the validation phase using logistic regression models with FTP as the response variable and parity score, BMI, breast density, HRT duration, OC use, and smoking history as predictors. Statistically significant P values are shown in bold.