Assess the empirical importance of Pearson centering

Determine how much Pearson centering of PnL series matters relative to transverse geometry and dispersion weighting when explaining the dependence between signal correlation and PnL correlation.

Background

The paper distinguishes three transformations connecting signal similarity to PnL dependence: dispersion weighting, temporal centering, and time-series normalization. In the special case of constant return dispersion, shared positive expected information coefficients can be removed by Pearson centering, leaving the correlation of the signals’ PnL innovations. The paper explicitly identifies the relative empirical importance of this centering effect, compared with transverse signal geometry and dispersion weighting, as unresolved.