Health indicators: eliminating bias from convenience sampling estimators.

Area of Science:

• Public Health
• Biostatistics
• Epidemiology

Background:

• Convenience samples (e.g., clinic visitors) are common in public health but introduce bias.
• Traditional random sampling avoids bias but can be inefficient and ignore valuable convenience data.
• Inferential limitations arise from convenience samples, hindering accurate population-level health indicator assessment.

Purpose of the Study:

• To develop a methodology for more accurate health indicator inference using combined data sources.
• To leverage both convenience and random sample data efficiently.
• To overcome the limitations of using convenience samples alone for public health inference.

Main Methods:

• A novel annealing methodology is proposed.
• This method integrates a small random sample with a larger convenience sample.
• The approach aims to balance data quality, cost-efficiency, and inferential power.

Main Results:

• The combined approach allows the use of powerful inferential tools typically reserved for random samples.
• The methodology provides more accurate health indicator information compared to using only random sample data.
• This integration overcomes the inherent biases of convenience samples.

Conclusions:

• Combining convenience and random samples via annealing offers a superior approach for public health inference.
• This method enhances the accuracy and utility of health indicator data.
• It presents a more efficient and informative strategy than relying solely on random sampling.