Update of Prior Probabilities by Minimal Divergence

Area of Science:

• Statistics
• Probability Theory
• Data Analysis

Background:

• Empirical probability distributions are fundamental in statistical modeling.
• Updating these distributions with new data is crucial for adapting models.
• Existing methods may not optimally balance new evidence with prior information.

Purpose of the Study:

• To investigate methods for updating empirical probability distributions.
• To compare updates that incorporate new observations and prior information.
• To explore the impact of different divergence measures on the update process.

Main Methods:

• Updating empirical probability distributions using new observational data.
• Employing minimization of Hellinger distance for optimal update.
• Employing minimization of quadratic Bregman divergence for optimal update.
• Considering updates incorporating conditional probability information.

Main Results:

• The update process successfully reproduces new observations while interpolating with prior information.
• Minimizing Hellinger distance and quadratic Bregman divergence yield different optimal updates.
• The choice of divergence measure influences the resulting probability distribution.

Conclusions:

• Both Hellinger distance and quadratic Bregman divergence provide valid frameworks for updating probability distributions.
• The divergence between these methods highlights the importance of selecting an appropriate measure.
• Further research can explore the implications of conditional probability updates.