Deterministic, quenched, and annealed parameter estimation for heterogeneous network models

Area of Science:

• Network analysis
• Statistical modeling
• Economic systems

Background:

• Two main statistical approaches exist for economic system analysis: econometrics and statistical physics.
• Recent work integrated these by minimizing Kullback-Leibler divergence, creating integrated and conditional models.
• Distinct parameter estimation methods are used in each approach.

Purpose of the Study:

• To compare different parameter estimation recipes for continuous, conditional network models.
• To determine the most effective estimation method by comparing econometric and statistical physics approaches.

Main Methods:

• The study compares deterministic, quenched, and annealed estimation methods.
• Focus is on continuous, conditional network models within an integrated framework.
• Analysis involves comparing parameter estimation strategies based on averaging and maximization orders.

Main Results:

• The annealed estimation recipe is identified as the best alternative to the deterministic one.
• This finding is specific to continuous, conditional network models.
• The study highlights the impact of averaging and maximization order on parameter estimation.

Conclusions:

• The annealed estimation method offers a more robust approach for analyzing economic networks compared to deterministic methods.
• This research provides valuable insights for selecting appropriate statistical models and estimation techniques in econometrics and network science.
• The findings contribute to the ongoing integration of statistical physics and econometric methodologies.