The ability of different imputation methods for missing values in mental measurement questionnaires

Area of Science:

• Psychological Measurement
• Data Science
• Statistical Analysis

Background:

• Incomplete data significantly impact mental measurement questionnaires, yet receive less attention than in clinical trials.
• Focusing on missing data handling in psychological research is crucial for accurate measurement.
• This study addresses the deficiency in handling missing questionnaire data.

Purpose of the Study:

• To compare the accuracy of four common methods for handling missing items in psychology questionnaires.
• To evaluate imputation methods based on varying non-response rates.
• To assess the performance of imputation techniques in psychological data.

Main Methods:

• Utilized datasets from self-acceptance scale (SAQ), activities of daily living scale (ADL), and self-esteem scale (RSES).
• Compared direct deletion, mode imputation, Hot-deck (HD) imputation, and multiple imputation (MI) using simulation and validation groups.
• Assessed methods using absolute deviation, root mean square error, and average relative error with missing proportions of 5-20%.

Main Results:

• Multiple imputation (MI) demonstrated the smallest bias across different missing proportions.
• Hot-deck (HD) imputation yielded the lowest absolute deviation in standard deviation values.
• Both MI and HD significantly outperformed direct deletion and mode imputation, with minimal differences in respondent scores compared to complete datasets.

Conclusions:

• Multiple imputation (MI) offers the best performance for handling missing data in psychological questionnaires, though it requires advanced analytical skills.
• Hot-deck (HD) imputation is a practical alternative for psychological investigations when MI is not feasible due to resource limitations.
• Both MI and HD imputation provide reliable results, significantly improving upon simpler methods like direct deletion and mode imputation.