Identification of the mean–scenario interaction

Determine whether the joint repair of the conditional mean and scenario channels exhibits genuine superadditivity, and quantify the conditional contribution and interaction magnitude, using adequately powered paired experiments that retain the necessary joint-versus-single-repair contrasts.

Background

The factorial experiment shows that jointly repairing the conditional mean and scenario channels improves Sharpe relative to the all-generator baseline, while the Shapley–Taylor estimate attributes a substantial point share to their interaction. Nevertheless, only six paired seeds were used, the confidence interval includes zero, and the relevant paired contrast between joint and scenario-only repairs was not retained. Thus, the existence and size of superadditivity remain unresolved.