Last updated: 2018-02-06

Code version: ee112bc

Introduction

In the Knockoff paper simulations, \(\beta\)’s are either \(0\) or \(A\). Here we are replicating the results, and investigating how well Knockoff deal with small signals.

In the following simulations, we always have \(n = 3000\), \(p = 1000\), \(X_{n \times p}\) has independent columns simulated from \(N(0, 1)\) and then normalized to have \(\|X_j\|_2^2 \equiv 1\). For a certain \(\beta\), \(Y_n \sim N(X_{n\times p}\beta_p, I_n)\). Out of \(p = 1000\) \(\beta_j\)’s, here are three scenarios.

  • Scenario 1: \(950\) are 0, \(25\) are \(3.5\), \(25\) are \(-3.5\). (replicating a data point on Fig 3 of the Knockoff paper)
  • Scenario 2: \(900\) are 0, \(100\) are \((N(0, 5.19^2) \wedge 3.5) \vee (-3.5)\) so that \(50\) of them are expected to be larger than \(3.5\) or smaller than \(-3.5\). And then we truncate these large signals to make them to be \(3.5\) or \(-3.5\).
  • Scenario 3: Similar to Scenario 2, only that \(800\) are 0, \(200\) are \((N(0, 3.04^2) \wedge 3.5) \vee (-3.5)\) so we still have on average \(50\) of them are expected to be \(3.5\) or \(-3.5\).

Scenario 1: 50 large signals, no small signals, 950 zeroes.

Scenario 2: 50 large signals, 50 small signals, 900 zeroes.

Scenario 3: 50 large signals, 150 small signals, 800 zeroes.

FDP.BH FDP.Knockoff Power.BH Power.Knockoff Power.Large.BH Power.Large.Knockoff Power.Small.BH Power.Small.Knockoff
0.0914 0.0623 0.4198 0.4398 0.4198 0.4398 NA NA
0.0880 0.0458 0.2970 0.2623 0.4748 0.4132 0.1210 0.1126
0.0754 0.0342 0.2216 0.1641 0.5206 0.3929 0.1236 0.0886

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 10.13.2

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     

other attached packages:
[1] ggplot2_2.2.1 knitr_1.19   

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.14     magrittr_1.5     munsell_0.4.3    colorspace_1.3-2
 [5] rlang_0.1.6      stringr_1.2.0    highr_0.6        plyr_1.8.4      
 [9] tools_3.4.3      grid_3.4.3       gtable_0.2.0     git2r_0.21.0    
[13] htmltools_0.3.6  yaml_2.1.16      lazyeval_0.2.1   rprojroot_1.3-2 
[17] digest_0.6.14    tibble_1.4.1     evaluate_0.10.1  rmarkdown_1.8   
[21] labeling_0.3     stringi_1.1.6    compiler_3.4.3   pillar_1.0.1    
[25] scales_0.5.0     backports_1.1.2

This R Markdown site was created with workflowr