Estimation and Calibration
Calibration chooses model parameters from observations; estimation also quantifies uncertainty. A close in-sample fit does not establish identification or predictive value.
Central questions
- Which likelihood, moments, or loss function connect data to parameters?
- Are parameters identifiable from available observations?
- How stable are estimates across time, instruments, and sampling choices?
Planned model
Fit a simple arrival or impact model to generated data while varying sample size, censoring, regime shifts, and misspecification.