JeffersonLab / japan

Just Another Parity ANalyzer
7 stars 13 forks source link

Feature: LRB matrix math #231

Closed wdconinc closed 5 years ago

wdconinc commented 5 years ago

Finally figured out how the matrix math expressions for the uncertainties on the slopes:

  dAxy.Zero();
  dAxy.Rank1Update(TMatrixDDiag(invRPP), TMatrixDDiag(mRYYp), norm); // diag mRYYp = row of ones
  dAxy.Sqrt();
  dAxy.NormByColumn(mSP); // divide
  dAxy.NormByRow(mSYp, ""); // mult

which is mean square error / ndof * R^-1. TMatrixD is designed for (RPN-like) in-place operations, hence the weird order of operations.

(Really, these were a few loose ends that weren't committed to feature-burst-regression before Paul merged that into develop.)