A Diagnostic Procedure for Detecting Outliers in Linear State-Space Models.

Area of Science:

• Statistics
• Data Analysis
• Psychometrics

Background:

• Outliers in longitudinal data pose unique challenges due to data correlation.
• Traditional methods often discard outliers, potentially losing valuable information.
• Distinguishing between additive and innovative outliers is crucial for accurate analysis.

Purpose of the Study:

• To develop and evaluate an outlier detection approach for multi-subject state-space models.
• To improve the power and reduce false detection rates of outlier identification.
• To demonstrate the practical application of the proposed method in real-world data.

Main Methods:

• A Monte Carlo simulation study was conducted to assess the performance of the outlier detection approach.
• The study adapted test statistics and detection procedures for different outlier types (additive vs. innovative).
• An open-source software implementation was developed for the proposed procedures.

Main Results:

• The proposed outlier detection approach demonstrated effective performance in identifying outliers in longitudinal state-space models.
• Adaptations led to improved statistical power and reduced false detection rates.
• The method proved useful in analyzing ecological momentary assessment data.

Conclusions:

• Outlier detection in longitudinal state-space models is essential for robust statistical inference.
• The developed approach offers a valuable tool for researchers dealing with complex longitudinal data.
• The findings highlight the importance of considering outlier characteristics for model improvement.