Semiparametric Particle Filters

• Activity Number: 502
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Carles Breto*
• Companies: Universidad Carlos III de Madrid
• Keywords: maximum likelihood ; state-space model ; iterated filtering ; particle filter

Abstract

Standard particle filters rely on a parametric specification for the measurement noise. I will present a residual-based approach where the distribution of the measurement noise is estimated non-parametrically via a prediction error decomposition. Such non-parametric estimate can be used in a second step to maximize the likelihood using plug-and-play approaches, like iterated filtering, which, instead of analytical results, only requires simulation from a numerical model. Such two-step approach can be used to improve the fit of non-linear non-Gaussian state-space models.