ASH vs Knockoff

Introduction

The true \(\beta\) are simulated as \(\beta \sim \pi_0\delta_0 + (1 - \pi_0)N(0, \sigma_\beta^2)\).

Independent design

\(X_{n \times p}\) has independent columns simulated from \(N(0, (1/\sqrt n)^2)\) so they are roughly normalized.

Local correlation design

\(X_{n \times p}\) has correlation \(\Sigma_{ij} = \rho^{|i - j|}\). Each row is independently \(N(0, \frac1n\Sigma)\).

Factor Model for \(X\)

Observation

1. Model-\(X\) knockoff is very powerful.
2. Using estimated distribution of \(X\) rather than the true distribution hurts the power of Model-\(X\) knockoff.
3. The power of Model-\(X\) knockoff using estimated distributio of \(X\) is on par with that of ASH and BH, probably because the presence of small signals makes knockoff less powerful.
4. Sometimes equi is better than SDP when generating knockoffs, as shown in previous simulations using factor model for \(X\).

Session information

sessionInfo()

R version 3.4.3 (2017-11-30)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.4

Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

loaded via a namespace (and not attached):
 [1] compiler_3.4.3  backports_1.1.2 magrittr_1.5    rprojroot_1.3-2
 [5] tools_3.4.3     htmltools_0.3.6 yaml_2.1.18     Rcpp_0.12.14   
 [9] stringi_1.1.6   rmarkdown_1.9   knitr_1.20      git2r_0.21.0   
[13] stringr_1.3.0   digest_0.6.15   evaluate_0.10.1