Maximum likelihood estimation for a special type of grouped data with an application to a dose-response problem

Area of Science:

• Biostatistics
• Statistical Modeling
• Pharmacometrics

Background:

• Maximum likelihood (ML) estimation is a standard statistical technique.
• Grouped data presents challenges for precise ML estimation.
• Existing methods may be computationally intensive for complex data structures.

Purpose of the Study:

• To develop a computationally efficient shortcut method for grouped maximum likelihood (ML) estimates.
• To apply this method to dose-response data with mixed grouping resolutions.
• To validate the method's performance on real and simulated datasets.

Main Methods:

• A novel algebraic approach for calculating grouped ML estimates.
• Application of the method to a modified quantal dose-response model.
• Utilizing data with varying degrees of grouping in different dimensions.

Main Results:

• The shortcut method provides accurate grouped ML estimates.
• The method is particularly useful for data with mixed grouping resolutions.
• Successful application to both simulated and real-world dose-response datasets.

Conclusions:

• The developed shortcut method is a practical and efficient tool for statistical analysis of coarsely and finely grouped data.
• This approach enhances the analysis of dose-response relationships where dosage levels are random and grouped.
• The method demonstrates robust performance, offering a valuable alternative for complex statistical modeling problems.