Outlier reduction by an option-3 measurement scheme

Area of Science:

• Medical Research
• Biostatistics
• Data Analysis

Background:

• Detecting changes in longitudinal data is crucial in medical research.
• Measurement outliers can significantly increase false alarm rates when identifying changes.
• Existing methods may be sensitive to outliers, compromising data integrity.

Purpose of the Study:

• To develop a novel method for detecting changes in longitudinal data robust to measurement outliers.
• To improve the accuracy of change detection in medical research by mitigating outlier effects.
• To demonstrate the efficacy of the new method using dental attachment probing data.

Main Methods:

• A new measurement scheme involving an adaptive number of repeated measurements is proposed.
• Initially, two measurements are taken; if within a threshold, their average is used.
• If measurements differ significantly, a third measurement is obtained, and the closest pair's mean is used.

Main Results:

• The proposed method demonstrates a considerable sample size advantage compared to naive repeated measurements.
• The scheme exhibits robustness against various outlier error distributions.
• The method effectively reduces the false alarm rate caused by measurement outliers.

Conclusions:

• The developed outlier detection method enhances the reliability of change detection in longitudinal medical data.
• This approach offers statistical advantages, including improved sample size efficiency and outlier robustness.
• The findings have practical implications for medical research, particularly in areas like dental attachment probing.