chapter  15
20 Pages

Time Series Analysis with a Skewed Kalman Filter

WithPhilippe Naveau, Marc G. Genton, and Caspar Ammann

The analysis of time series has always played a central role in statistics. In this chapter, we focus on temporal state-space models for time series that can be decomposed in a sum of three components: a smooth trend, a skewness part, and noise. The first one is supposed to be deterministic and the other two random. Such time series decompositions are common in paleoclimate studies. The central element of the estimation procedure is the Kalman filter that has to be modified to deal with skewness.