[Estimation of median regression from measured values]

Area of Science:

• Statistics
• Econometrics

Background:

• Median regression offers an alternative to mean regression, particularly for non-normally distributed data.
• Understanding the properties and estimation of median regression is crucial for robust statistical modeling.

Purpose of the Study:

• To define and demonstrate the median regression function.
• To prove a lemma regarding the continuity and differentiability of the median regression function.
• To develop and compare methods for estimating the median regression function.

Main Methods:

• Definition and examples of the median regression function.
• Proof of a lemma for continuity and differentiability.
• Estimation via empirical distribution function.
• Estimation via kernel estimation with Gaussian kernels.
• Comparison with empirical regression of the first kind.

Main Results:

• Sufficient conditions for continuity and differentiability of the median regression function are established.
• Two distinct methods for estimating the median regression function from sample data are derived.
• Demonstration of estimation techniques using GALTON's historical data.
• Comparison highlights differences between median and mean-based regression estimates.

Conclusions:

• The median regression function is well-defined and estimable using both empirical and kernel-based approaches.
• The derived lemma provides theoretical underpinnings for the function's properties.
• The study offers practical insights into applying median regression for data analysis and curve fitting.