Fig. 3.

HTRX is accurate and necessary for Logistic Regression. Comparison of the average out-of-sample R² through 10-fold CV for linear and logistic regression models in a simulated dataset. Feature set specifies the maximum number of features (of 6) that can interact using ‘SNP’, ‘2SNP_hap’, ‘3SNP_hap’, etc., and ‘all_hap’ represents the all the possible haplotypes, while ‘HTR’ uses templates that interact between all SNPs with no ‘X’ in the template. $D=50\%$, B = 10 and q = 3 are used for ‘Two-stage-CV’ (Algorithm 2). ‘Direct-Fit’ refers to all-feature multivariate regression. ‘Direct CV’ is the implementations of Algorithm 1. (a) Linear Models. (b) Logistic Regression Models