Module econometrics 

Econometrics: OLS, WLS, 2SLS, logistic MLE, GMM, and calibration records.

Allocation class: HotZeroHeap. All scratch uses fixed-capacity stack arrays. No Vec/String/Box in any kernel.

Assumptions:

• OLS assumes exogeneity (E[ε|X] = 0), iid errors, no perfect multicollinearity. Standard errors are not yet computed (future work).
• WLS assumes known weights proportional to inverse error variance.
• 2SLS assumes instrument relevance (n_instr >= n_reg) and exogeneity. Underidentified models (n_instr < n_reg) are refused.
• Logistic MLE assumes iid Bernoulli outcomes with logit link; uses Newton-Raphson (IRLS).

Structs

CalibrationRecord

A repr(C) calibration record linking a fitted model to its data and diagnostics.

Enums

EconometricsError

Constants

MAX_OBSERVATIONS

Maximum observations in a bounded regression.

MAX_REGRESSORS

Maximum regressors (including constant) in a bounded regression.

Functions

gmm_moment_eval

Evaluate GMM moment conditions: g(theta) = (1/n) sum m_i(theta).

iv_2sls_into

Two-Stage Least Squares (2SLS).

logistic_mle_into

Logistic regression via Newton-Raphson (IRLS).

ols_into

Ordinary Least Squares via normal equations X'X b = X'y.

wls_into

Weighted Least Squares: (X'WX) b = X'Wy.