Version 3.1.3

Fixed crash on stat(c) after OLS

Version 3.1.2

Incorporated small-sample factor in r(dist)

Version 3.1.1

Minor bug fixes and speed-ups

Version 3.1.0

Radical speed-up of Wild Restricted Efficient bootstrap after IV/2SLS/LIML/k-class. Now works after ivreg2 when using that command's partial() option. Dropped support for linear GMM. Altered algorithm for searching for confidence interval bounds, slightly affecting CI results. Added unit tests.

Version 3.0.2

Dropped "KK" calculation (last expression in eq 60 in paper) because inefficient when interpolating. Refined plotting to minimize interpolation anchor resets. Refined criterion for using "granular"-optimized code (many small clusters).

Version 3.0.1

Compiled it in Stata 13, not Stata 11. Problem was causing crashes.

Version 3.0.0

Much more fully exploits linearity discussed in appendix A.2 of Roodman, MacKinnon, Neilsen, and Webb (2019), bringing 10X speed gain when forming confidence interval for Wald test after OLS and Anderson-Rubin test after IV.

Version 2.8.1

Fix 2.8.0 bugs. Even more fully exploit linearity, for ~2X speed gain in inverting test after OLS.

Version 2.8.0

More fully exploit linearity of numerators and K matrices as in appendix A.2, for modest speed gain. Recenter classical score test even for non-OLS, reversing 2.4.1 change. Add ptolerance() option.

Version 2.7.3

Prevented crash on WRE bootstrap-t with svmat(numer)