Statistical validation of event predictors: A comparative study based on the field of seizure prediction

Journal Article · Wed Jun 15 00:00:00 EDT 2011 · Physical Review. E, Statistical, Nonlinear, and Soft Matter Physics (Print)

DOI:https://doi.org/10.1103/PHYSREVE.83.066704· OSTI ID:21554536

Feldwisch-Drentrup, Hinnerk ^[1]; Schulze-Bonhage, Andreas ^[1]; Timmer, Jens ^[1]; Schelter, Bjoern ^[2]

---

1. Bernstein Center Freiburg (BCF), University of Freiburg, Freiburg (Germany)
2. Freiburg Center for Data Analysis and Modeling (FDM), University of Freiburg, Freiburg (Germany)

The prediction of events is of substantial interest in many research areas. To evaluate the performance of prediction methods, the statistical validation of these methods is of utmost importance. Here, we compare an analytical validation method to numerical approaches that are based on Monte Carlo simulations. The comparison is performed in the field of the prediction of epileptic seizures. In contrast to the analytical validation method, we found that for numerical validation methods insufficient but realistic sample sizes can lead to invalid high rates of false positive conclusions. Hence we outline necessary preconditions for sound statistical tests on above chance predictions.

• OSTI ID: 21554536
• Journal Information: Physical Review. E, Statistical, Nonlinear, and Soft Matter Physics (Print), Vol. 83, Issue 6; Other Information: DOI: 10.1103/PhysRevE.83.066704; (c) 2011 American Institute of Physics; ISSN 1539-3755
• Country of Publication: United States
• Language: English

Related Subjects

71 CLASSICAL AND QUANTUM MECHANICS
GENERAL PHYSICS
46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND TECHNOLOGY
COMPARATIVE EVALUATIONS
COMPUTERIZED SIMULATION
MONTE CARLO METHOD
PERFORMANCE
SOUND WAVES
STATISTICS
VALIDATION
CALCULATION METHODS
EVALUATION
MATHEMATICS
SIMULATION
TESTING