Tutorial: Assessing the impact of nonignorable missingness on regression analysis using Index of Local Sensitivity to...

Area of Science:

• Psychology
• Statistics
• Data Science

Background:

• Missing data is prevalent in psychological research, often necessitating the assumption of data missing at random (MAR).
• The MAR assumption is unverifiable and incorrect application can introduce bias into statistical analyses.
• Assessing the robustness of findings to potential violations of the MAR assumption is crucial for credible research.

Purpose of the Study:

• To introduce a novel class of sensitivity analyses for evaluating the impact of departures from the MAR assumption.
• To present the Index of Local Sensitivity to Nonignorability (ISNI) as a practical measure of robustness.
• To provide an accessible R package (isni) for implementing these sensitivity analyses in regression models.

Main Methods:

• The study derives sensitivity analyses from the Index of Local Sensitivity to Nonignorability (ISNI).
• ISNI offers a computationally straightforward approach, avoiding complex non-MAR missing-data models.
• The method is implemented in the R package 'isni' for various regression models.

Main Results:

• The proposed method, ISNI, provides a computable measure to assess the sensitivity of conclusions to non-MAR data.
• The 'isni' R package simplifies the application of these sensitivity analyses.
• Illustrative analyses on real-world psychology datasets demonstrate the practical utility of the method.

Conclusions:

• The ISNI method offers a valuable tool for psychologists to conduct sensitivity analyses and assess the credibility of their findings under potential MAR violations.
• The accompanying R package makes advanced missing data sensitivity analysis accessible to researchers.
• This approach enhances the rigor and trustworthiness of psychological research employing statistical modeling.