Penalized multiple inflated values selection method with application to SAFER data.

Area of Science:

• Statistics
• Biostatistics
• Epidemiology

Background:

• Count data often exhibit multiple values with higher frequencies than expected under standard models.
• Existing multiple-inflated Poisson models rely on complex manual selection of inflated values, risking missed data points.
• Accurate modeling of such data is crucial in fields like clinical trials and public health.

Purpose of the Study:

• To develop and validate a novel, automated two-stage method for selecting inflated values in count data.
• To improve the performance of the multiple-inflated Poisson model in parameter estimation and inflated value identification.
• To demonstrate the generalizability of the proposed method using real-world clinical trial data.

Main Methods:

• A two-stage inflated value selection process is proposed, treating all count response values as potential inflated points.
• Adaptive lasso regularization is employed on the mixing proportions to identify significant inflated values.
• The method's performance is evaluated through numerical studies and simulations based on clinical trial data.

Main Results:

• The proposed two-stage method demonstrates excellent performance in both selecting inflated values and estimating model parameters.
• Numerical studies confirm the method's superiority over existing approaches in identifying true inflated points.
• Simulations based on HIV intervention trial data show the method's robustness and applicability to real-world scenarios.

Conclusions:

• The novel two-stage inflated value selection method offers a more efficient and accurate approach for analyzing multiple-inflated count data.
• This method enhances the utility of the multiple-inflated Poisson model, particularly in complex datasets from clinical research.
• The approach is validated for practical application, promising improved insights from epidemiological and clinical studies.