Factor Analysis and Structural Equation Modeling of Actigraphy Derived Sleep Variables

Abstract

Actigraphy is a useful tool in the measurement of sleep, with many applicable variables that indicate the quality of a study participant’s sleep. Typical actigraphy derived variables from sleep studies include sleep efficiency, sleep latency, total sleep time, and the number of nightly awakenings. Actigraphy can also give parameters for autocorrelation and sinusoidal fitting that are used in circadian rhythm modeling. All of these variables are useful in inferential statistics to determine differences between groups. However, no work has been done that examines a set of relationships among these various sleep variables, which can possibly reduce the dimension of the set into a smaller number of interpretable latent factors, which can aid in characterizing sleep quality. Factor analyses and structural equation modeling were used to assess how these measured sleep variables load into different factors. Analyses showed that these variables load into 3 factors: sleep time, sleep activity and circadian rhythm.