A two-step approach to simultaneously correct for selection and misclassification bias in nonprobability samples from...

Area of Science:

• Statistics
• Public Health
• Sociology

Background:

• Nonprobability sampling is common for hard-to-reach populations but prone to selection bias.
• Existing data integration methods assume accurate probability sample data, which may not hold due to misclassification bias.

Purpose of the Study:

• To propose a novel two-step statistical approach to address misclassification bias in probability samples.
• To enhance data integration methods for nonprobability samples by correcting errors in external probability samples.
• To accurately estimate prevalence for minority populations, such as sexual minority men.

Main Methods:

• A two-step statistical method was developed to correct misclassification bias in probability samples.
• Simulated data were used to evaluate the method's performance under varying misclassification rates.
• A bootstrap variance estimator was validated for the proposed approach.

Main Results:

• The proposed method significantly reduced bias in data integration.
• The bootstrap variance estimator proved effective across low, moderate, and high misclassification rates.
• The approach is particularly beneficial when misclassification rates are high.

Conclusions:

• The novel two-step method effectively addresses misclassification bias, improving data integration for nonprobability samples.
• This approach enhances the reliability of prevalence estimates for sensitive characteristics in population surveys.
• The study demonstrates the method's utility in estimating social connectedness among sexual minority men.