Principal Component Analysis
PCA finds orthogonal directions of maximum variance in a covariance (or correlation) matrix — the yield-curve level/slope/butterfly and many equity ‘statistical factors’ are PCA.
Definition
Principal Component Analysis refers to the yield-curve level/slope/butterfly and many equity ‘statistical factors’ are PCA. Keep that definition fixed when comparing series, managers, or regimes — renaming the same tape does not create a new signal.
Why it matters
It is a named object desks use to frame risk, positioning, or process. When the yield-curve level/slope/butterfly and many equity ‘statistical factors’ are PCA shifts, related hedges, limits, and narratives usually need an explicit update rather than a quiet assumption.
Case
Suppose a desk is positioned for the opposite of what principal component analysis is saying. If the yield-curve level/slope/butterfly and many equity ‘statistical factors’ are PCA moves against that book, the first question is not “is the story clever?” but whether size, hedges, and stop logic still match the observation.
How to read it
Keep the definition fixed, then challenge it with cross-checks before sizing. Prefer a short written null hypothesis for Principal Component Analysis: what would falsify the current reading in the next window?