A Robust Rerank Approach for Feature Selection and Its Application to Pooling-Based GWA Studies

Figure 4

Performance comparison of different methods for feature selection with equal sample size in two groups. The 𝑥-axis represents the number of selected SNPs (𝑀₁) based on simulated DNA pooling data. The 𝑦-axis represents the average number of identified truly relevant SNPs over 100 simulations. The shaded area represents the possible results of SAM with various 𝑠₀. The 𝑠₀̂ value is automatically determined by samr R package.

(a) (𝑛₀,𝑛₁)=(16,16)

(b) (𝑛₀,𝑛₁)=(8,8)

(c) (𝑛₀,𝑛₁)=(16,16) with pooling error

(d) (𝑛₀,𝑛₁)=(8,8) with pooling error

(e) (𝑛₀,𝑛₁)=(16,16) with outliers

(f) (𝑛₀,𝑛₁)=(8,8) with outliers