Simulated Data Example
The report below comes from a connectome with a known amount of confounding,
generated by the SEM-based simulator and analysed with confound control on both
axes (selection_input="residualized", model_input="residualized"), at
Schaefer100-17 size so the brain figures are included.
For the runnable code, start from the
regression quickstart — it uses the same simulator and shows
nested cross-validation for the p-threshold. For a conceptual guide to the
simulator (the generative model, the edge classes, and the R2 / kappa knobs),
see Simulating Data; for the demonstration that partial
selection alone does not deconfound, see scripts/confound_inflation_demo.py.
The report
Note the headline, which names the confound configuration the report was produced under — one report is one cell of the 2×2. See Interpreting Results for a guide to every section.