Public articles linked to the same research event.
arXiv The study proposes a model-agnostic Bayesian state-space post-processing framework that treats overlapping forecasts of the same future time issued from different origins as noisy, potentially biased measurements of a common latent trajectory, enforcing cross-origin coherence without access to the forecasting model's architecture, parameters, or training data; across 100 simulation replications it reduced cross-origin incoherence by a mean 58.5% (SD 15.2%) with mean latent-trajectory correlation 0.994, and in an S&P 500 volatility application it moved 90% prediction interval coverage from 94.8% to 88.3% and lowered CRPS by 8.7% while RMSE rose about 15%, with LSTM forecasts already showing low raw incoherence that reconciliation eliminated.
The study proposes a model-agnostic Bayesian state-space post-processing framework that treats overlapping forecasts of the same future time issued from different origins as noisy, potentially biased measurements of a common latent trajectory, enforcing cross-origin coherence without access to the forecasting model's architecture, parameters, or training data; across 100 simulation replications it reduced cross-origin incoherence by a mean 58.5% (SD 15.2%) with mean latent-trajectory correlation 0.994, and in an S&P 500 volatility application it moved 90% prediction interval coverage from 94.8% to 88.3% and lowered CRPS by 8.7% while RMSE rose about 15%, with LSTM forecasts already showing low raw incoherence that reconciliation eliminated.
The study proposes a model-agnostic Bayesian state-space post-processing framework that treats overlapping forecasts of the same future time issued from different origins as noisy, potentially biased measurements of a common latent trajectory, enforcing cross-origin coherence without access to the forecasting model's architecture, parameters, or training data; across 100 simulation replications it reduced cross-origin incoherence by a mean 58.5% (SD 15.2%) with mean latent-trajectory correlation 0.994, and in an S&P 500 volatility application it moved 90% prediction interval coverage from 94.8% to 88.3% and lowered CRPS by 8.7% while RMSE rose about 15%, with LSTM forecasts already showing low raw incoherence that reconciliation eliminated.
The study proposes a model-agnostic Bayesian state-space post-processing framework that treats overlapping forecasts of the same future time issued from different origins as noisy, potentially biased measurements of a common latent trajectory, enforcing cross-origin coherence without access to the forecasting model's architecture, parameters, or training data; across 100 simulation replications it reduced cross-origin incoherence by a mean 58.5% (SD 15.2%) with mean latent-trajectory correlation 0.994, and in an S&P 500 volatility application it moved 90% prediction interval coverage from 94.8% to 88.3% and lowered CRPS by 8.7% while RMSE rose about 15%, with LSTM forecasts already showing low raw incoherence that reconciliation eliminated.