Bayesian inference for misaligned irregular time series with application to palaeoclimate reconstruction

Abstract

This thesis proposes new Bayesian methods to jointly analyse misaligned irregular time series. Temporal misalignment occurs wdien multiple irregularly spaced time series are considered together, or when the time periods defining the data points are not the same across different series. Other issues under consideration include errors in the time scales, and non-Gaussian processes for underlying latent values. Our proposed models are hierarchical.