Statistical Methods¶
This section documents the mathematical foundations behind MORIE’s estimators. The methods are dataset-agnostic — they apply to any suitably-shaped tabular input, including the OTIS placement records, TPS incident feeds, CPADS survey data, and any other dataset that matches the estimator’s signature.
- Causal Estimands
- Causal Inference
- Potential outcomes framework
- Inverse Probability Weighting (IPW) — Hájek estimator
- Average Treatment Effect on the Treated (ATT)
- Average Treatment Effect on the Controls (ATC)
- Augmented IPW (AIPW) — Doubly Robust
- G-computation (Outcome Regression)
- eBAC-selection-adjusted IPW
- Sensitivity Analysis
- Average Treatment Effect on the Treated (ATT)
- Average Treatment Effect on the Controls (ATC)
- Group Average Treatment Effect (GATE)
- Conditional Average Treatment Effect (CATE)
- Local Average Treatment Effect (LATE / IV)
- Interactive Regression Model (IRM)
- References
- Propensity Scores
- Double Machine Learning (DML)
- eBAC — Estimated Blood Alcohol Concentration
- Survey-Weighted Statistics
- Survey Sampling
- Dataset-Agnostic Analysis
- Psychometric Methods
- Ontario Restrictive Confinement (OTIS)
- OTIS Linkage Constraints — Read Before Doing Any Individual-Level Analysis
- MRM modules
- SIU IAP — Federal Structured Intervention Unit Implementation Advisory Panel
- Sprott-Doob CRIMSL + Schulich Law SIU analyses
- Doob T-539-20 Federal Court affidavit replication
- Toronto Police Service (TPS) Statistics
- Hawkes Self-Exciting Point Processes
- Spatial Statistics
- Statistical Physics of Crime
- Key Empirical Findings
- TurboQuant — Vector Quantization
- MORIE Inference Engine
- Signal Processing & Biomedical Analysis
- Homomorphic Deconvolution & Cepstral Analysis
- Post-Quantum Cryptography (Research)
- Population Genetics
- Polyglot REPL
- Deployment
Quick reference¶
Each entry below names the estimator, the estimand it targets, and the Python function that produces it.
Causal estimators¶
run_propensity_ipw_analysis— IPW (Hájek), ATEestimate_att— IPW (Hájek), ATTestimate_atc— IPW (Hájek), ATCestimate_aipw— AIPW (doubly robust), ATEestimate_gate— GATE (AIPW per group)estimate_cate— T-learner / S-learner, CATE (per unit)estimate_late— 2SLS / Wald IV, LATEestimate_ate— DML–PLR, ATEestimate_irm— DML–IRM, heterogeneous ATEestimate_pliv— DML–PLIV, LATEestimate_ate_gcomputation— G-computation, ATErun_ebac_selection_ipw_analysis— eBAC-IPW, selection-adjusted ATEe_value— E-value sensitivity boundsensitivity_rosenbaum— Rosenbaum bounds
Logistic / model comparison¶
run_weighted_logistic_analysis— weighted logistic, ORcompare_nested_logistic_models— nested-model LRT
Survey + descriptive statistics¶
morie.surveyhelpers — survey-weighted CIs and prevalencehorvitz_thompson_total— HT estimator, population totalhajek_mean— Hájek estimator, population mean
Power + Bayes¶
run_power_design_module— N required for a given designBeta-binomial Bayes — posterior mean / CI (see
morie.causal)
Psychometrics¶
crba— Cronbach’s α (internal consistency)mcdo— McDonald’s ω (reliability)kmo— KMO sampling adequacybart— Bartlett’s sphericity (factorability)paran— Parallel analysis (factor retention)crel— Composite reliabilityave— Average Variance Extracted (convergent validity)
OTIS-specific¶
rplace— Regional placement counts / proportionsastcmb— Alert-state combination encodingvolat— Regional volatility (movement metric)otdml— DML IRM on correctional data, ATE / ATT