A Class of Robust Estimators for Moment Condition Models

Area of Science:

• Statistics
• Econometrics

Background:

• Moment condition models are widely used but sensitive to outliers and misspecification.
• Existing minimum empirical divergence estimators can be unreliable under data contamination.

Purpose of the Study:

• Introduce a novel class of robust estimators for moment condition models.
• Provide robust alternatives to standard minimum empirical divergence estimators.
• Address the sensitivity of estimation procedures to outliers and model deviations.

Main Methods:

• Construct estimators using truncated orthogonality functions.
• Minimize divergences in their dual form.
• Analyze influence functions and prove robustness and consistency theoretically.

Main Results:

• The proposed estimators limit the impact of outliers and model deviations.
• Theoretical proofs confirm the robustness and consistency of the new estimators.
• Monte Carlo simulations demonstrate that extreme observations do not disproportionately affect the estimates.

Conclusions:

• The novel robust estimators offer improved reliability for moment condition models.
• These estimators provide a valuable tool for statistical and econometric analysis in the presence of outliers.
• The findings support the use of these robust methods for more stable parameter estimation.