Interactive explainer

Interactive: Causal inference: separating effects from associations

Causal inference asks how an outcome would change under an intervention. An observed difference between groups answers that question only under additional assumptions about common causes and comparable treatment options.

Interactive

Interactive: Causal inference: separating effects from associations

Change training allocation, the true effect and the share of difficult cases. Every value comes from a fully specified synthetic model.

Reading the display

The first bar compares the groups as actually composed. The second weights easy and difficult cases using the same target-population proportions for both groups. The table reports success rates with and without training and the probability of receiving training.

“Training for difficult cases” also sets easy cases’ training probability to its complement: 80% for difficult cases means 20% for easy ones. Allocation changes without silently changing the true effect. The display uses expected proportions without sampling noise; fractional counts are not observed people.

Three experiments

  1. Simpson’s paradox: inspect the default state. The raw comparison is negative even though training helps within both groups.
  2. Random allocation: set the first control to 50%. Both difficulty groups receive training equally often, and raw and adjusted differences coincide.
  3. Another target population: change the share of difficult cases. Group averages change while the standardised effect still equals the effect specified in this model.

Reset restores 80% training allocation, a ten-percentage-point effect and 50% difficult cases. Real data may contain unobserved common causes. Adjustment is correct here because the model is known. The causal inference article explains the assumptions required before applying that interpretation to observations.

From NyxAI

smartlytics is the learning hub of NyxAI GmbH.

Explore the methods we work with, explained openly, with sources and calculations you can reproduce.

Decision Core

NyxAI is developing a Decision Core for Decision Intelligence that combines optimisation, simulation, statistics and reinforcement learning: compare options, make uncertainty visible and learn from outcomes, with people in control.

AI workflows at NyxAI

DeepSquare

Explainable chess AI with human playing styles: NyxAI’s testing ground for understandable decisions under time pressure.