Skip to main content

Table 1 Summary of TDR analysis for binary genotypes and covariates

From: Application of two machine learning algorithms to genetic association studies in the presence of covariates

 

Approach

Model

Include Z

Stratify by Z

Residualize by Z

Ignore Z

1. ADDITIVE

+/+

+/+

+/+

+/+

2. ADDITIVE WITH CONFOUNDING

++/+

+/+

+/+

++/++

3. ADDITIVE WITH EFFECT MEDIATOR

- -/- -

- -/- -

- -/- -

-/-

4. INTERACTION WITH MAIN EFFECTS

++/++

++/++

++/++

++/++

5. INTERACTION WITH NO MAIN EFFECTS

+/++

+/++

-/+

-/+

6. CONDITIONAL ASSOCIATION

+/+

+/+

-/+

-/+

  1. Summary of simulation results in Figures 1 and 2: Results are given in pairs corresponding to RF and MARS respectively; a "+" indicates reasonable TDR (≥ 80%) for detecting moderate effect sizes (≥ 0.5); "++" indicates reasonable TDR (≥ 80%) for detecting small effect sizes (≥ 0.3);"-" indicates lower TDR (50% to 80%) for moderate effect sizes; and '--" indicates very low TDR (< 20%) at moderate effect sizes. Correlation between X1 and Z is fixed at 0.5 for MODEL 2.