Nearest neighbor matching for bivariate outcomes?

Hi all,

I'm new to nearest neighbor matching. I've been using it to analyze data with continuous outcomes; can this technique also be used for bivariate outcomes? I've seen some papers switch from nnmatch to propensity score (psmatch) with bivariate outcomes, but I have not seen an explicit discussion of nnmatch with bivariates.

Hmm for clarification are you using nearest neighbor matching for case/control collection or are you using Nearest Neighbor Analysis for cases? Sorry but nearest neighbor case control selection is so common that's what I think of when you say "matching" but nearest neighbor analysis is a different algorithm altogether.

The only way I can see it being used for bivariate is a case of stepwise regression which eliminates all other variables as significant within the regression but just thinking about it makes me say "no". Then again you can always do a propensity score to attenuate the bivariate if all other covariates are sensibly nonsignificant. Depends on your data and your knowledge of what you're investigating.

Out of curiosity what is it you're studying to use NN analysis? If it's genomic data then be prepared to enter the Biostatistics Boxing Ring.