Modelling record times in sport with extreme value methods

We exploit connections between extreme value theory and record processes to develop inference methods for record processes. Record processes have trends in them even when the underlying process is stationary. We study the problem of estimating the underlying trend in times achieved in sports from record data. We develop new methods for inference, simulating record series in non-stationary contexts, and assessing fit which account for the censored characteristic of record data and we apply these methods to athletics and swimming data