Overview
Michael K. Pitt is an econometrician (University of Reading / King's College London) working on simulation methods for state-space models, particle filtering, and stochastic volatility. With Neil Shephard he developed the auxiliary particle filter and likelihood-based methods for non-Gaussian time series.
Key Contributions / Features
- Auxiliary particle filter — Pitt–Shephard (1999): With Neil Shephard, introduced the auxiliary-variable particle filter that resamples with a look-ahead at the next observation, curing the bootstrap filter's weight degeneracy under informative/outlying data. See Particle Filter and Pitt-Shephard (1999).
- Non-Gaussian measurement time series — Shephard–Pitt (1997): MCMC likelihood analysis of non-Gaussian state-space models via block sampling of the states.
- Factor stochastic volatility — Pitt–Shephard (1999b): Time-varying covariances via a factor stochastic-volatility model.
Related