Identification of units placed into matched sets

Given a bipartite matching (object of class optmatch), create a logical vector of the same length indicating which units were and were not placed into matched sets.

Usage

matched(x)

unmatched()

matchfailed(x)

Arguments

x: Vector of class optmatch (especially as generated by a call to fullmatch).

Value

A logical vector (without names).

Details

matched and unmatched indicate which elements of x do and do not belong to matched sets, as indicated by their character representations in x.

When fullmatch has been presented with an inconsistent combination of constraints and discrepancies between potential matches, so that there exists no matching (i) with finite total discrepancy within matched sets that (ii) respects the given constraints, then the matching problem is said to be infeasible. TRUEs in the output of matchfailed indicate that this has occurred.

Note

To understand the output of matchfailed element-wise, note that fullmatch handles a matching problem in three steps. First, if fullmatch has been directed to match within subclasses, then it divides its matching problem into a subproblem for each subclass. Second, fullmatch removes from each subproblem those individual units that lack permissible potential matches (i.e. potential matches from which they are separated by a finite discrepancy). Such "isolated" units are flagged in such a way as to be indicated by unmatched, but not by matchfailed. Third, fullmatch presents each subproblem, with isolated elements removed, to an optimal matching routine. If such a reduced subproblem is found at this stage to be infeasible, then each unit contributing to it is so flagged as to be indicated by matchfailed.

Author

Ben Hansen

Examples

data(plantdist)

mxpl.fm0 <- fullmatch(plantdist) # A feasible matching problem
#> Warning: Without 'data' argument the order of the match is not guaranteed
#>     to be the same as your original data.
c(sum(matched(mxpl.fm0)), sum(unmatched(mxpl.fm0)))
#> [1] 26  0
sum(matchfailed(mxpl.fm0))
#> [1] 0
mxpl.fm1 <- fullmatch(plantdist, # An infeasible problem
                      max.controls=3, min.controls=3)
#> Warning: Without 'data' argument the order of the match is not guaranteed
#>     to be the same as your original data.
#> Warning: Matching failed. (Restrictions impossible to meet?)
#>  Enter ?matchfailed for more info.
c(sum(matched(mxpl.fm1)), sum(unmatched(mxpl.fm1)))
#> [1]  0 26
sum(matchfailed(mxpl.fm1))
#> [1] 26

mxpl.si <- factor(c('a', 'a', 'c', rep('d',4), 'b', 'c', 'c', rep('d', 16)))
names(mxpl.si) <- LETTERS[1:26]
mxpl.exactmatch <- exactMatch(mxpl.si, c(rep(1, 7), rep(0, 26 - 7)))
# Subclass a contains two treated units but no controls;
# subclass b contains only a control unit;
# subclass c contains one treated and two control units;
# subclass d contains the remaining twenty units.
# only valid subproblems will be used

mcl <- c(1, Inf)

mxpl.fm2 <- fullmatch(plantdist + mxpl.exactmatch,
                      max.controls=mcl)
#> Warning: Without 'data' argument the order of the match is not guaranteed
#>     to be the same as your original data.
sum(matched(mxpl.fm2))
#> [1] 22

table(unmatched(mxpl.fm2), matchfailed(mxpl.fm2))
#>        
#>         FALSE
#>   FALSE    22
#>   TRUE      1

mxpl.fm2[matchfailed(mxpl.fm2)]
#> factor(0)

mxpl.fm2[unmatched(mxpl.fm2) &   # isolated units return as
         !matchfailed(mxpl.fm2)] # unmatched but not matchfailed
#>    I 
#> <NA>