Analysis with missing data in drug prevention research

Area of Science:

• Statistics
• Prevention Research
• Data Analysis

Background:

• Missing data is a persistent challenge in prevention research.
• Statistical solutions exist but are underutilized in practice.
• This chapter bridges the gap between statistical theory and applied prevention research.

Purpose of the Study:

• To introduce systematic application of modern missing data analysis techniques.
• To provide practical guidance for handling missing data in prevention studies.
• To address specific issues like respondent burden, attrition, and costly measurement.

Main Methods:

• Focuses on missing data analysis for continuous, normally distributed data.
• Applicable to analyses using covariance matrices, particularly within the general linear model.
• Illustrates methods with examples from drug prevention research.

Main Results:

• Recommends Expectation-Maximization (EM) algorithm or other maximum likelihood procedures (e.g., multiple imputation).
• Highlights that appropriate methods maximize use of available data without generating new information.
• Emphasizes that alternative analyses may produce biased results.

Conclusions:

• Researchers should prioritize EM algorithm or multiple imputation for missing data.
• Alternative analyses should be used cautiously due to potential bias.
• Understanding and modeling the cause of missingness is crucial.
• Adjusting estimates for originally missing sampled cases can improve accuracy.