Carving as randomization | Data | Select on half of the data with glmnet | Inference | Coverage over repeated experiments

Setup | Data and fit | Inference | For a gaussian response | Coverage over repeated experiments | Reference

Randomization and the proportion used for selection | Data | Fit the LASSO on the randomized response | Inference | Coverage over repeated experiments

The selection event in two coordinates | Data: a wide design | Fit the LASSO on the randomized data | Screen the top 10 inactive variables | Ignoring the screening step | The LASSO-solution basis | First-dropped versus exact conditioning | Coverage over repeated experiments

The swapped coefficient as a contrast | Setup | Swap estimands | Example 1: independent features | Does screening matter? | Example 2: correlated features and surrogates