Performance of WCU-S vs WCU-C in score-variance tests

Determine whether the WCU-S wild cluster bootstrap provides superior finite-sample performance than the WCU-C variant when implementing score-variance tests for deciding the appropriate level of clustering in linear regression models, and quantify the conditions under which such improvement holds.

Background

Score-variance tests can be implemented using wild cluster bootstrap resampling. The mnwsvt package currently uses the classic unrestricted WCU-C bootstrap. The author suggests that the score-based WCU-S bootstrap may perform even better, but this has not been established. A systematic comparison would guide best practice for clustering-level diagnostics.