Adjusted logistic propensity weighting methods for population inference using nonprobability volunteer-based...

Area of Science:

• Epidemiology
• Biostatistics
• Population Health

Background:

• Nonprobability volunteer-based samples are common in epidemiologic studies due to practical constraints.
• These samples limit the ability to make accurate finite population (FP) inferences because they lack representativeness.
• Existing inverse propensity score weighting methods struggle with accurate participation rate estimation.

Purpose of the Study:

• To develop a novel method for estimating participation rates in nonprobability samples.
• To enable more reliable population-level inferences from volunteer-based studies.
• To provide an efficient and implementable weighting strategy for epidemiologic research.

Main Methods:

• Proposed an adjusted logistic propensity weighting (ALP) method to estimate nonprobability sample unit participation rates.
• Incorporated scaling of survey sample weights to enhance estimator efficiency.
• Developed Taylor linearization variance estimators to account for all sources of variability in FP mean estimation.

Main Results:

• The ALP method produces approximately unbiased estimators for population quantities, irrespective of the nonprobability sample rate.
• The method is easily implemented using available software.
• Empirical evaluation using NHANES III and NHIS data demonstrated the method's utility in estimating mortality rates.

Conclusions:

• The adjusted logistic propensity weighting (ALP) method offers a robust solution for drawing population inferences from nonprobability samples.
• ALP provides a practical and statistically sound approach to address sampling biases in volunteer-based health research.
• This method enhances the utility of readily available nonprobability survey data for public health surveillance and analysis.