An R package for optimal record adjustment
The package adjusts numeric records x, such that Ax <= b is met after adjustment. The adjustment is kept 'as small as possible' by minimizing (x-x0)'W(x-x0), where x0 is the original value, x the final value and W a diagonal matrix with positive weights. Here, x may be fairly large; rspa has been tested with vectors of about half a million elements under 60k (in)-equality restrictions.
The main addition I'd like to make is to be able to use a general (sparse) W-matrix.