In sadeh's paper(1), in which he proposed an algorithm for actigraphy based sleep/wake classification, the threshold for probability of sleep parameter is 0 (hence, greater than 0 => epoch is considered sleep, and awake otherwise).
However, I saw several implementations of the algorithm online, where a threshold of -4 is used instead of 0.
My question is why lowering the threshold to -4 ? Actigraphy tends already to overestimate sleep,,,, lowering the threshold has to worsen this tendency... Can you explain to me why ? Thank you
Reference:
(1) “Activity-Based Sleep-Wake Identification: An Emprical Test of Methodological Issues”, Avi Sadeh, K.M. Sharkey and M. A. Carskadoon, Sleep17(3): 201-207, 1994 American Sleep Disorders Association and Sleep Research Society