When cleaning physiological sensor data, what is the primary drawback of using simple mean imputation compared to time-series interpolation methods?
The primary drawback of using simple mean imputation for physiological sensor data is the loss of temporal dependency, which occurs because mean imputation replaces missing values with the global average of the dataset regardless of the surrounding data points. Physiological signals, such as heart rate or blood oxygen levels, are time-series data, meaning each measurement is statis....
Community Answers
Sign in to open profiles and full community answers.
Lara Emad Aljudaibi
“The primary drawback of using simple mean imputation for psychological sensor data is the loss of temporal dependency, which occurs because mean imputation replaces missing values with the global average of the dataset regardless of the surrounding data points. Physiological signals, such as heart rate or blood oxygen levels, are time-series data, meaning each measurement is statistically related to the measurements taken immediately before and after it. When you use simple mean imputation, you effectively inject a static, unrelated value into a dynamic sequence, which artifically flattens the variance and destroys the underlying patterns or trends of the signal. In contrast, time-series interpolation methods, such as linear interpolation, estimate missing values based on the slope between the existing data points immediately preceding and following the gap. This preserved the local trend and continuity of the signal, ensuring the imputed values remain consistent with the psychological state of the subject at that specific moment. Consequently, mean imputation introduces significant bias into the data that can cause downstream analytical models to misinterpret the signal's rhythm or rate of change, whereas interpolation maintains the signal's temporal coherence.”
100.0%