Characterizing the performance of the Conway-Maxwell Poisson generalized linear model

Area of Science:

• Risk Analysis
• Statistical Modeling
• Probability Distributions

Background:

• Count data are common in risk analysis, including deaths, accidents, and system failures.
• Traditional regression models for count data have limitations in variance structure.
• The Conway-Maxwell Poisson (COM-Poisson) distribution offers a more flexible alternative.

Purpose of the Study:

• To assess the performance of the maximum likelihood estimation (MLE) for the COM-Poisson generalized linear model (GLM).
• To characterize the parameter estimation accuracy of the COM-Poisson GLM.
• To estimate the prediction accuracy of the COM-Poisson GLM using simulated data.

Main Methods:

• Fitting the COM-Poisson GLM using maximum likelihood estimation (MLE).
• Utilizing simulated data sets to evaluate model performance.
• Assessing parameter estimation and prediction accuracy.

Main Results:

• The COM-Poisson GLM effectively models under-, equi-, and overdispersed count data.
• Accurate parameter estimates were achieved using the MLE implementation.
• The model demonstrated strong prediction accuracy on simulated data.

Conclusions:

• The COM-Poisson GLM is a flexible and accurate tool for count data regression in risk analysis.
• It overcomes limitations of traditional models by handling various dispersion types.
• This approach shows promise for improved risk assessment through robust count data modeling.