I have previously done Nearest-neighbor propensity score matching in R using the MatchIt package, where I had 20 patients in the treatment group, and 100 in the control group, and found 20 matching controls.

Now, I have 15 patients in one group, and 16 in another. Can I find matching pairs in this, such that just the patients which are comparable are included, and the rest are dropped. So for e.g. I get 10 matched pairs?

I will be using RStudio for this.



1 Answer 1


I'm not sure what you're hoping to accomplish with 10 patients per group. That is way too small for any meaningful analysis. Any bias will be dramatically overpowered by the variance. Only in very unusual cases can quantitative analysis be supported with such a tiny sample. This might be a good situation to switch to a qualitative analysis in which you learn as much as you can from your patients through interviews rather than trying to compare means on some variable.

All that said, if you do want to go through with this, you can do so by applying a caliper, which is a limit on how distant a pair of units can be. By applying a tighter and tighter caliper, you can drop more and more pairs until only as many as you want remain. You can place calipers on the propensity score or on any variable you are matching on. The MatchIt package makes this easy; just use the caliper argument, which is documented in help("matchit").


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.