Dealing with missing data in multi-informant studies: A comparison of approaches

Area of Science:

• Social and behavioral sciences
• Psychometrics
• Quantitative psychology

Background:

• Multi-informant studies are prevalent but present analytical challenges due to shared/unique informant data and incompleteness.
• Handling incomplete data from multiple sources requires robust statistical methods.

Purpose of the Study:

• To compare the performance of three methods for analyzing incomplete multi-informant data.
• To evaluate approaches considering reference and nonreference informants.
• To identify the most effective method for handling missing data in multi-informant research.

Main Methods:

• Monte Carlo Simulation was employed to compare analytical approaches.
• Three methods were examined: two-method measurement model for planned missing data (2MM-PMD), auxiliary variable approaches (FIML/MI), and listwise deletion.
• The study simulated incomplete multi-informant datasets with varying missingness patterns.

Main Results:

• The two-method measurement model for planned missing data (2MM-PMD) demonstrated the best performance.
• 2MM-PMD showed superior accuracy in point estimates, Type I error rates, and statistical power when data were missing at random.
• This method also exhibited greater robustness when data were not missing at random.

Conclusions:

• The two-method measurement model for planned missing data (2MM-PMD) is recommended for analyzing incomplete multi-informant data.
• This approach offers improved accuracy and robustness compared to auxiliary variable methods and listwise deletion.
• Proper specification of 2MM-PMD is crucial for optimal results in social and behavioral research.