How to predict the probability of a nuclear accident using past observations? What increase in probability the Fukushima Dai-ichi event does entail? Many models and approaches can be used to answer these questions. Poisson regression as well as Bayesian updating are good candidates. However, they fail to address these issues properly because the independence assumption in which they are based on is violated.
We propose a Poisson Exponentially Weighted Moving Average (PEWMA) based in a state-space time series approach to overcome this critical drawback. We find an increase in the risk of a core meltdown accident for the next year in the world by a factor of ten owing to the new major accident that took place in Japan in 2011.