Answers are grounded in this paper and cite their sources.
Financial prediction intervals must respond to volatility clustering and abrupt shifts in forecast-error distributions. We propose regime-gated feedback adaptive conformal inference (RG-ACI), which combines volatility-normalized nonconformity scores, rolling calibration, a single online miscoverage-rate state and a causal short-to-long volatility-ratio gate. Building on the finance-specific score normalization of Chen et al., the method couples a retrospective coverage correction with an ex-ante volatility signal without introducing a second error-feedback state. On 20 simulated heavy-tailed, autoregressive stochastic-volatility paths with multiple regime transitions, RG-ACI attains 0.895 mean coverage and 0.894 crisis-period coverage at a nominal 90% target with low between-path dispersion, while volatility-normalized rolling calibration remains the most efficient interval in width and interval score. Component ablations separate the effects of error-rate adaptation, the volatility gate and feedback gain, and quantify the calibration–efficiency tradeoff of each. The contribution is an explicitly causal, single-state integration of feedback and regime gating together with a transparent analysis of its tradeoffs under designed distribution shifts.
No takes yet. Share an insight, caveat, or question.
Hayes et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: