Skip to main content
. Author manuscript; available in PMC: 2021 Apr 30.
Published in final edited form as: HGG Adv. 2021 Jan 5;2(2):100019. doi: 10.1016/j.xhgg.2020.100019

Table 5.

Mean prediction performance in METS of pairwise model intersecting genes

Population EN versus RF EN versus SVR EN versus KNN
AFA 0.1075 versus 0.1021 (p = 2.8 × 10−3) 0.1072 versus 0.0857 (p = 2.2 × 10−20) 0.1111 versus 0.0691 (p = 6.4 × 10−6)
HIS 0.0793 versus 0.0960 (p = 2.1 × 10−11) 0.0797 versus 0.0799 (p = 0.95) 0.0846 versus 0.0616 (p = 1.8 × 10−13)
CAU 0.0555 versus 0.0699 (p = 1.1 × 10−8) 0.0535 versus 0.0588 (p = 0.024) 0.0592 versus 0.0450 (p = 3.7 × 10−6)
AFHI 0.0991 versus 0.0924 (p = 8.7 × 10−6) 0.0975 versus 0.0815 (p = 9.9 × 10−19) 0.1058 versus 0.0661 (p = 1.7 × 10−49)
ALL 0.0909 versus 0.0871 (p = 0.017) 0.0902 versus 0.0774 (p = 2.8 × 10−14) 0.1041 versus 0.0628 (p = 1.5 × 10−50)

We performed paired t tests between the prediction performance of EN and each of the other machine learning models with each training population. The t test p values are shown in parentheses. EN, elastic net; RF, random forest; SVR, support vector regression; KNN, K nearest neighbor; AFA, MESA African American; HIS, MESA Hispanic American; CAU, MESA European American; AFHI, MESA African American and Hispanic American; ALL, all MESA.