Multiply robust imputation procedures for zero-inflated distributions in surveys.

Area of Science:

• Statistics
• Survey Methodology
• Data Science

Background:

• Survey data often suffers from item nonresponse, where respondents omit answers.
• Standard single imputation methods may be inadequate, particularly for variables with many zero values.
• Addressing missing data is crucial for accurate statistical inference.

Purpose of the Study:

• To propose novel multiply robust imputation procedures for survey variables with a high proportion of zero values.
• To enhance the accuracy and robustness of statistical estimates in the presence of item nonresponse.
• To provide a reliable method for handling complex missing data patterns in surveys.

Main Methods:

• Development of multiply robust imputation procedures based on multiple imputation and/or nonresponse models.
• Definition of multiply robustness: estimator consistency even if all but one model are misspecified.
• Utilizing a generalized jackknife procedure for estimating the variance of imputed estimators.

Main Results:

• Simulation studies indicate the proposed procedures perform well.
• Demonstrated effectiveness in reducing bias and improving efficiency of estimates.
• Showcased adequate coverage rates for estimated parameters.

Conclusions:

• Multiply robust imputation offers a powerful approach for handling survey data with many zero values.
• The proposed methods provide a statistically sound and reliable alternative to traditional imputation techniques.
• These procedures enhance the integrity of survey data analysis, leading to more trustworthy results.