Establishing a threshold for the number of missing days using 7 d pedometer data.

Area of Science:

• Physical Activity Measurement
• Data Imputation Techniques
• Biostatistics

Background:

• Accurate measurement of physical activity, often using pedometers, is crucial for health research.
• Missing data in pedometer datasets is a common challenge that can affect study validity.
• The individual information (II)-centered approach offers a method for imputing missing data.

Purpose of the Study:

• To determine the maximum number of missing daily recovery days that can be imputed using the individual information (II)-centered approach without significantly impacting results.
• To evaluate the accuracy of the II-centered approach for imputing varying amounts of missing pedometer data.

Main Methods:

• Utilized data from 86 participants with 7 days of complete pedometer wear (Yamax SW 200).
• Simulated 1 to 5 missing days of data 10,000 times, imputing values with the II-centered approach.
• Compared imputed datasets to complete datasets using repeated measures ANOVA and calculated Mean Absolute Percentage Error (MAPE).

Main Results:

• No significant differences (p > 0.05) were found between the complete dataset and datasets with 1 to 5 imputed missing days.
• Mean Absolute Percentage Errors (MAPE) were lowest for 1 day (5.2%) and 2 days (8.4%) of missing data.
• MAPEs exceeded 10% for datasets with 3 or more imputed missing days.

Conclusions:

• The individual information (II)-centered approach can effectively impute up to two days of missing pedometer data without significant alteration of results.
• Imputing more than two days of missing step count data using this method is not recommended due to increased error.
• These findings provide guidance for handling missing data in pedometer-based physical activity research.