Maximum likelihood based analysis of equally spaced longitudinal count data with first-order antedependence and...

Area of Science:

• Statistics
• Biostatistics
• Longitudinal Data Analysis

Background:

• Longitudinal count data often exhibit over-dispersion, where variance exceeds the mean.
• Standard Poisson models may be inadequate, leading to biased results.
• Accurate modeling of over-dispersion is crucial for reliable statistical inference.

Purpose of the Study:

• To develop and implement a maximum likelihood approach for longitudinal count data with over-dispersion.
• To compare the proposed method with existing approaches that ignore over-dispersion.
• To provide accessible R code for practical application.

Main Methods:

• Maximum likelihood estimation tailored for equally spaced longitudinal count data.
• Simulation studies to evaluate method performance under various conditions.
• Application to real-world datasets: seizure data and German Socio-Economic Panel data.

Main Results:

• The proposed maximum likelihood method effectively accounts for over-dispersion.
• Simulations show improved small-sample efficiency and correct coverage probabilities compared to methods ignoring over-dispersion.
• The approach demonstrates practical utility in analyzing complex datasets.

Conclusions:

• Accounting for over-dispersion in longitudinal count data is essential for accurate statistical analysis.
• The developed maximum likelihood method provides a robust and efficient alternative.
• The availability of R code facilitates the adoption of this method in diverse research areas.