Dynamic Generator Inversion for Observable Conditional Distributions

This project develops a distributional econometric framework for systems that move through discrete adjustments but are evaluated through observable outcomes such as prices, quantities, inventories, choices, durations, or stress states.

The central object is a conditional adjustment generator: a law that maps the current state of a system into future events or transitions. Doi-Peliti notation is used as a representation language. Evaluating the generator’s Hamiltonian at imaginary momentum gives a conditional characteristic exponent; derivatives give conditional cumulants; and a validation-selected regularized empirical-characteristic-function estimator links the generator to observable predictive distributions.

The working paper treats familiar models such as vector autoregressions, Gaussian transition densities, volatility equations, Hawkes specifications, and quantile regressions as projections, anchors, or benchmarks for the same distributional forecasting problem. Its evidence is currently diagnostic and reproducible rather than a final empirical market claim.

Working paper PDF (non-peer-reviewed).