nearest neighbor matching

[benthestatistician]

A fast, scalable function that, given an ISM, matches each treatment group member to its nearest neighbor (or k nearest neighbors) and outputs an optmatch object understood by optmatch handlers, would have various uses:

• users looking to consider full or pair matches alongside of NN matches could do both things within optmatch, using optmatch utilities/workflows to effect the comparison
• Ratios of control:treatment as realized in the NN match might be taken as a starting or ending point in an exploration of full matches with various mean.control= parameter values
• summary.InfinitySparseMatrix() could use it to assemble its summaries of the distribution of nearest neighbors. (The current tapply()-based approach doesn't scale well to large problems.)