Description

1️⃣ Classical PCA minimizes squared error (L₂ norm) — symmetric, mean-based.
2️⃣ This study extends PCA to asymmetric norms, targeting tail behavior via expectiles (L₂τ,₂).
3️⃣ Expectiles generalize means and quantiles; they capture asymmetric risk and heavy-tail dynamics.
4️⃣ Define Principal Expectile Components (PECs): maximize τ-variance (Var_τ(X)) or minimize asymmetric loss.
5️⃣ PEC = eigenvectors of a weighted covariance matrix (C_τ) with τ-dependent weights.
6️⃣ LAWS algorithm (asymmetric weighted least squares) iteratively updates expectiles and weights to convergence.
7️⃣ PECs reduce to classical PCA for elliptical distributions — ensuring interpretability and consistency.
8️⃣ Simulation studies show PECs outperform classical PCA in skewed, heteroscedastic, heavy-tailed settings.
9️⃣ Applications include fMRI risk-attitude data and Chinese temperature extremes, revealing asymmetric climate trends.
🔟 PECs provide a robust, distribution-aware alternative to PCA — bridging expectile regression and dimension reduction for tail-risk analysis.

Would you like me to turn this into a Quantinar-ready 10-slide outline (each bullet as one slide with image/formula cues)?