ABSTRACT
The predictability of a system depends on how the collected data are measured and thus is a function of the technology used to make it. This applies to both physical and stochastic models. This chapter provides information to determine the predictability of seizures. Data that consist of only the times at which epileptic events start were collected from different patients spanning different intervals of time. These events look random, but may in fact be correlated. Determining whether this system has memory, that is, identifying the presence of long-range dependence, could mean that these events are predictable. The fact that memory exists at all in epilepsy, beyond very short times, is a positive discovery: at least seizures could be predictable. If no memory was observed then no clear evidence exists that seizures are at all predictable. Because event times are not random, information about the timing of a current and past seizure does contain information about future seizures.
