halfmoon (development version)
ess()is now a re-export of the generic of the same name from causalgenerics. Attaching halfmoon alongside another package that re-exports that same generic no longer produces a masking conflict, because both packages export the one object. A package that defines its own unrelatedess()still masks, as before. The calculation is unchanged for numeric weights.Because the generic names its first argument
x,ess()no longer accepts the argument namewts. Pass the weights positionally, as iness(w).ess(), andbal_ess()through it, now error on non-numeric input instead of returning a meaningless number. Previouslyess(NULL)andbal_ess(NULL)returnedNaN, and factors, logicals, data frames, dates, time differences, and complex vectors each produced a value:bal_ess(factor("a"))returned1.8.ess(rep(0, 5))andess(numeric(0))still returnNaN.
