Skip to main content

Table 4 Overview of main results for the primary estimands of our simulation study

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

Method

Stability

Coverage

Power

Type 1 error

Lasso_CV_Split

No concern

Acceptable

Low

Acceptable

Lasso_CV_PoSI

No concern

Too high

Low

Low

Lasso_CV_SI

Problematic

Acceptable

Acceptable

High

Lasso_Neg_SI

Problematic

Acceptable

Acceptable

Acceptable

ALasso_CV_Split

No concern

Acceptable

Acceptable

Acceptable

ALasso_CV_PoSI

No concern

Too high

Acceptable

Acceptable

ALasso_CV_SI

Problematic

Too low

High

High

ALasso_Neg_SI

Problematic

Acceptable

High

High

  1. By “Acceptable” we mean that results were mostly (i.e. in median over all scenarios) within the expected simulation variability