Helper: recursive forecasts (WG procedure) | Table 1: five tests across three OOS specifications | WG Figure 1: cumulative SSE difference for DP | Takeaway | References

Setup | Replication | Results | References

Data | Methodology | Out-of-sample loop | Results | Key spca_est() features used | References

Overview | Quick start | PCA, PLS, and RRA | Scaled PCA | IPCA | Prediction | Factor evaluation | Bundled datasets

The IPCA model | Data preparation | Fitting IPCA | Validation against the Python ipca package | Multiple factors | References

The data | Replicating Table A2 | References

Recursive forecasts | ENC-NEW across lag orders | Comparison with Clark-West MSFE-adjusted | References

Per-horizon picture | Uniform multi-horizon SPA | Average multi-horizon SPA | Custom weights | Block-length sensitivity | Takeaways | References

The data | Forecasting setup | Replicating the OOS DM panel of Table 1 | AR(1) | AR(2) | Single-country deep dive: Japan, AR(1), recursive | References

Overview | Pairwise Forecast Comparison | Diebold-Mariano test | Clark-West test | Giacomini-White test | Multiple Forecast Comparison | Hansen's SPA test | CSPA test | Visualization | CSMS: confidence set for the most superior | Predictive Regressions | IVX-Wald test | Elliott-Muller test | References

CEPA across forecast horizons | Choice of instruments | Note on the original paper | References

Setup | SPA test against the primary RV proxy | Robustness across realised-variance proxies | Top 10 alternatives by mean loss (5-min proxy) | References

Figure 2: JNJ, one-versus-one CSPA | Figure 3: JNJ, one-versus-all CSPA | Table 4 Panel A: cross-stock rejection counts | S&P 500: confidence set for the most superior method | Takeaway | References

Helpers | Technical indicators | Takeaway | References

Helpers | Macro predictors | Takeaway | References