Table 3 Mathematical approaches to quantifying variability of mood data

From: Reporting guidelines on remotely collected electronic mood data in mood disorder (eMOOD)—recommendations

Statistical method Assumptions Limitations
Time domain e.g., RMSSD Normally distributed data Influenced by extreme scores No estimate of the width of the distribution Do not distinguish different signals Examples of datasets with identical means, SDs and RMSSDs with very different underlying data structure
Frequency domain/ Spectral analysis Data considered a sum of sinusoidal oscillations with distinct frequencies Analyses require stationarity within data Long data series required
Entropy Considered a measure of randomness/irregularity Should be calculated on non- normalized time series Accuracy reduced in short time series Sinusoidal trends are detrimental Spikes in the data can impair linear estimates