frameworkDecisions and computation

Causal inference and counterfactuals

Distinguish prediction and association from claims about what would happen under an intervention.

Working definition

Causal inference asks how an outcome would differ under alternative interventions, using a declared causal graph, identification assumptions, and study design. Randomization can identify effects under compliance and measurement conditions; observational analyses require stronger, testable and untestable assumptions.

Notation

ATE = E[Y(1) − Y(0)]Y ⟂ T | X under conditional exchangeability

Assumptions

  • Treatment, outcome, and intervention are well defined.
  • Confounders required for identification are addressed.
  • Interference and selection are considered.

Invariants

  • Association alone does not identify intervention effect.
  • Adjustment follows the causal graph rather than predictive importance.
  • A counterfactual contrast requires a target population.

Reproducible procedure

  • Draw the assumed causal structure.
  • Choose a design and identification strategy.
  • Estimate effects with falsification and sensitivity analyses.

Error and boundary controls

  • Unmeasured confounding may dominate.
  • Positivity failures prevent comparison.
  • Measurement and selection bias can reverse estimates.

What this does not establish

Unless celestial timing is manipulated or otherwise identified under a defensible design, predictive association must not be described as celestial causation.

Explicit applications

1 cross-domain bridges

Empirical validationempirical test

Prediction-versus-causation boundary

Specify when a benchmark can support predictive skill and why it usually cannot identify celestial causation.

Inputs

  • study design
  • assignment mechanism
  • outcomes and covariates

Outputs

  • supported claim type
  • unidentified paths
  • sensitivity analysis

Transformation: Map identification assumptions and test observable implications.

Limit: Most observational celestial-timing studies can test incremental prediction, not physical causation.

Open connected system →

Authoritative references

  1. [1]NIST/SEMATECH e-Handbook of Statistical Methods · National Institute of Standards and Technology

    Establishes: Methods for uncertainty analysis, calibration, time-series modeling, process monitoring, experimental design, reliability, and statistical comparison.

    Boundary: Statistical procedures quantify evidence under a design and model; they do not repair biased sampling, outcome leakage, post-hoc hypotheses, or unmeasured confounding.

Related mathematical concepts