Changed the var argument in plot(., type = "extrapolation") to variables (var still works as an abbreviation).
Added a unit testing framework.
Weights now have an average of 1 in each treatment group. This only affects the display of the weights, making it easier to assess extreme weights, but does not affect balance or any other property of the weights.
When using cluster-robust standard errors in lmw_est(), the degrees of freedom for the corresponding t-testing as now computed as one less than the number of clusters to be consistent with other software.
"IPWRA" is now an acceptable alias for "WLS" when supplied to the dr.method of lmw() with base weights. IPWRA is the name Stata gives to the same estimator.
When a weightit or matchit object is supplied to obj in lmw() or lmw_iv(), the focal component of the supplied object is automatically supplied and does not need to be specified. A warning will be thrown if focal is supplied and differs from the focal component of the supplied object.
Errors are prettier.