Application of a Hierarchical Model to Paleoenvironmental Time Series with Latent Times

• Activity Number: 354
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Aaron Springford* and David J. Thomson
• Companies: Queen's University at Kingston and Queen's University at Kingston
• Keywords: Time series ; Chronology model ; Hierarchical ; Bayesian ; Computation ; Spectrum estimation

Abstract

Observation times in time series data are usually assumed to be known and evenly spaced. These assumptions enable many methods for statistical inference of the measured process, such as ARMA models and spectral estimation using the Fast Fourier Transform. Thankfully, the measurement process is usually controlled by the experimenter, limiting the effect of these assumptions by design. However, for paleoenvironmental studies based on core data, measurement times must be inferred from depth and dating information (using radioisotopes, for example). Inference of measurement times is required for depths with and without dating information. Challenges include: (1) the requirement of monotonicity (increasing age with depth); (2) the ability to handle complicated likelihoods and to be robust to errors in dating estimates; (3) the propagation of timing uncertainty into time series analysis; and (4) expeditious computation. Contemporary approaches, which model core deposition, have difficulty resolving records without prior information that is often unavailable. We present an integrated hierarchical approach to address all four challenges that does not require informative prior information.