Finite-sample estimation

[buddejul]

Goal: Estimation of intervals from data using tolerance approach suggested in MST 2018 ECMA (see equation (27)).

Key idea: All functions of the data currently computed from the DGP have to be estimated

Todo:

• Estimate parameters and primitives from the data
• Note: Previously just computed identified parameters using the DGP, now need to estimate them directly (e.g. via Wald-Estimator)
• Relax constraint in AMPL program to only hold up to tolerance

Questions:

• Are these all just plug-in estimators? --> Estimate all relevant quantities up-front and just plug into existing functions
  • f_z, prop_z, covariances and variances