balanceTest() should work without explicit data argument

[benthestatistician]

data(nuclearplants)
with(nuclearplants, xBalance(pr~cap+ pt))

Doesn't work. It may be that some minor adjustments would get it to work, namely:

1. Changing R/xBalance.R:L2 from

data,

over to

data=NULL,

and
2. inserting a switch around R/xBalance.make.stratwts.R:L31-36 so that, in the event that data is NULL, sweights is evaluated as

        do.call(swt.ls[[nn]],
                args=list(data=data.frame(Tx.grp=zz[goodstrat.df[[nn]]],

                              stratum.code=factor(ss.df[goodstrat.df[[nn]],nn])

                              )
),
                envir=parent.frame())

ie, the code ,data[goodstrat.df[[nn]],,drop=FALSE] is removed.

[benthestatistician]

Cross-linking w/ #15, which may be easier pickings once this can of worms has been opened. (And vice versa.)

[benthestatistician]

in light of d8bf73c , this may be working now on the clusters branch. Remains to be tested.

[benthestatistician]

I think this is basically enabled for balanceTest(), but w/ some gaps. E.g., the following doesn't work:

  set.seed(20180207)
  n <- 100
  x1 <- rnorm(n)
  x2 <- rnorm(n)
  x3 <- 0.5 + 0.25 * x1 - 0.25 * x2 + rnorm(n)
  y <- sample(rep(c(1,0), n/2), prob = exp(x3) / (1 + exp(x3)))

  xy <- data.frame(x1, x2, x3, y)
  wts <- rpois(n, 7)
  balanceTest(y ~ x1 + x2 + x3, data = xy, unit.weights = wts)

rather you have to do

xy.wts <- data.frame(xy, wts)
balanceTest(y ~ x1 + x2 + x3, data = xy.wts, unit.weights = wts)

oughta fix this.

[benthestatistician]

The counterexample in my last comment seems to have been fixed (in proj1-balT) somewhere along the way. OTOH

> balanceTest(y ~ x1 + x2 + x3)
Error in terms.formula(formula, data = data) : 
  argument is not a valid model