Nonlinear State-Space Econometrics for Trading Signals With Python: Particle Filters, SMC², and Rao–Blackwellization for Real-Time Trading Signals (Richman Computational Economics) - Softcover

Inhaltsangabe

Level up your quant edge with a dense, practitioner-first playbook to design, estimate, and deploy nonlinear state-space models for live trading. From heavy-tailed returns and microstructure noise to high-dimensional factor SV and regime switching, you’ll master particle methods, SMC², and Rao–Blackwellized filters—then implement them line-by-line in Python.

What makes this the go-to resource

• Trading-first focus: Every model is motivated by alpha, risk, execution, and portfolio constraints.
• Real-time ready: Online filtering, fixed-lag smoothing, and latency-aware pipelines for production.
• Non-Gaussian by default: Robust heavy tails, jumps, count/intensity models, and discrete regimes.
• Scales with your universe: Factor SV, conditional independence, and GPU-friendly parallelism.
• Variance reduction that matters: Rao–Blackwellization, guided proposals, and tempered SMC for sharp likelihoods.

How each chapter delivers value

• Theory: Clear, mathematically precise derivations tailored to financial use-cases.
• Checkpoint MCQs: Multiple-choice questions with solutions to cement understanding quickly.
• Full Python code: End-to-end, reproducible demos for filtering, smoothing, PMCMC, SMC², and RBPF.

You will learn to

• Build robust nonlinear state-space models for alpha, volatility, liquidity, and execution costs.
• Engineer observation models for heavy tails, jumps, and microstructure distortions.
• Implement bootstrap/APF filters, guided proposals, and backward-simulation smoothers.
• Train via Particle EM, PMMH, Particle Gibbs/PGAS, and nested SMC².
• Collapse linear–Gaussian substructures with Rao–Blackwellization for speed and accuracy.
• Evaluate and select models with evidence estimates, prequential scoring, and DMA.
• Deploy GPU-accelerated pipelines with reproducibility and numerical stability.

Who this is for

• Quant researchers and portfolio managers seeking deployable signal pipelines.
• Data scientists and ML engineers moving beyond static models to state-space systems.
• Grad students in econometrics/finance looking for a rigorous, hands-on guide.

What you’ll build in code

• RBPF for dynamic regression with stochastic volatility and heavy tails.
• SMC² for online parameter learning across multi-asset universes.
• PGAS for regime-switching and semi-Markov duration models.
• Tempered SMC for evidence estimation and model comparison.
• Real-time signal extractors with risk forecasts (VaR/ES) and transaction-cost-aware P&L.

Die Inhaltsangabe kann sich auf eine andere Ausgabe dieses Titels beziehen.