Extreme value theory for stochastic integrals of Legendre polynomials (Q1006679)

The author studies the extremal behavior of stochastic integrals of Legendre polynomial transforms with respect to Brownian motion. As the main results, the author obtains the exact tail behavior of the supremum of these integrals taken over intervals \([0,h]\) with \(h>0\) fixed, and the limiting distribution of the supremum on intervals \([0,T]\) as \(T\rightarrow \infty \). One shows further how this limit distribution is connected to the asymptotic of the maximally selected quasi-likelihood procedure that is used to detect changes at an unknown time in polynomial regression models. In an application to global near-surface temperatures, it is demonstrated that the limit results presented in this paper perform well for real data sets.