Prediction model-based kernel density estimation when group membership is subject to missing.

Area of Science:

• Statistics
• Data Analysis
• Biostatistics

Background:

• Estimating density functions is crucial for understanding heterogeneous populations.
• Kernel smoothing is a common nonparametric method for density estimation.
• Missing subpopulation membership complicates standard density estimation.

Purpose of the Study:

• To develop new kernel smoothing methods for density function estimation with missing subpopulation membership.
• To address limitations of existing methods under the missing completely at random (MCAR) assumption.
• To accommodate the missing at random (MAR) assumption in density estimation.

Main Methods:

• Proposed novel kernel smoothing techniques incorporating prediction models for missing membership.
• Developed asymptotic properties for the new density estimators.
• Utilized simulation studies and a real-world mental health dataset for evaluation.

Main Results:

• The new kernel smoothing methods demonstrate improved performance in density function estimation.
• The proposed methods are effective under the missing at random (MAR) assumption.
• Performance was validated through simulations and a mental health study.

Conclusions:

• The novel kernel smoothing methods provide a robust approach for density estimation in the presence of missing data.
• These methods extend the applicability of kernel smoothing beyond the MCAR assumption.
• The findings have implications for analyzing complex population data in various fields, including mental health research.