Efficient testing of earthquake forecasting models

摘要: Computationally efficient alternatives are proposed to the likelihood-based tests employed by Collaboratory for Study of Earthquake Predictability assessing performance earthquake likelihood models in forecast testing centers. For conditional L-test, which consistency catalogue with a model, an exact test using convolutions distributions is available when number earthquakes period small, and central limit theorem provides approximate large. Similar methods R-test, compares likelihoods two competing models. However, like N-test fundamentally data model. We propose alternative test, based on classical paired t-test, more directly compare Although predicated normality assumption, this new T-test not computer-intensive, easier interpret than becomes increasingly dependable as increases.