Skip to main content

Table 1 Simulation setup

From: Evaluating methods for Lasso selective inference in biomedical research: a comparative simulation study

  

Toy setup

Realistic setup

Fixed design parameters

Motivation

Simplicity, insight

Realistic data

Number of variables

4

17

Type of variables

Continuous

Continuous, binary

Distribution of variables

Gaussian

Mixed

(Supplementary Table S2)

Varying design parameters

Correlation structures \({\varvec{\Sigma}}\)

7 blocked correlation matrices with no or strong correlation (Supplementary Material Sect. 3.2)

Fixed, mimicking real study

(Supplementary Material Sect. 3.3)

Coefficient structures \({\varvec{\beta}}\)

10

(Supplementary Table S1)

13

(Supplementary Table S3)

True target R2 (noise \({\varvec{\epsilon}}\))

0.2, 0.5, 0.8

0.2, 0.5, 0.8

Observations per variable

5, 10, 50

5, 10, 50

Simulation parameters

Number of scenarios

630

117

Iterations per scenario

900

900

  1. See the description of the simulation setups in the Supplementary Material for more details on the correlation and coefficient structures. Varying design parameters are the parameters that were varied in the full factorial design in both setups