Implementation and applications of bootstrap methods for the National Immunization Survey

Area of Science:

• Statistics
• Survey Methodology
• Biostatistics

Background:

• Complex probability sample surveys involve numerous weight adjustments to minimize potential bias in estimates.
• Adjustments include sampling design (SD) and non-sampling design (NSD) weight adjustments, addressing sampling plans and non-sampling errors, respectively.
• Existing variance estimates often incorporate SD adjustments but rarely all NSD adjustments, potentially biasing estimates and affecting confidence interval coverage.

Purpose of the Study:

• To implement and illustrate the bootstrap method for accounting for both SD and NSD weight adjustments in complex survey data.
• To evaluate standard confidence intervals that do not fully account for NSD adjustments.
• To obtain confidence intervals for ranks and assess logistic regression predictive power using receiver operating characteristic (ROC) curve analyses.

Main Methods:

• The study implements the bootstrap method to address weight adjustments in complex survey data.
• Utilizes data from the National Immunization Survey (NIS) for practical illustration.
• Applies bootstrap for evaluating confidence intervals, rank estimation, and logistic regression predictive power via ROC analysis.

Main Results:

• The bootstrap method provides a means to account for both SD and NSD weight adjustments in complex survey data.
• Demonstrates the potential bias in variance estimates and confidence intervals when NSD adjustments are omitted.
• Illustrates the application of bootstrap for enhanced analysis of survey data, including rank estimation and predictive modeling.

Conclusions:

• The bootstrap method is a valuable tool for improving the accuracy of variance estimates and confidence intervals in complex survey data.
• Properly accounting for all weight adjustments (SD and NSD) is crucial for reliable survey estimates.
• The bootstrap facilitates robust statistical inference and model evaluation in the context of complex survey designs.