use caliper but get completely different results than last version about matchit HOT 3 CLOSED

Yes, the behavior when using calipers has changed. To replicate old results, you must use an older version of MatchIt. That said, if you did not set a seed when you originally ran the analysis, your results will not replicate anyway.

In the past, matchit() would randomly select one control unit within the caliper of a treated unit. Now, matchit() selects the closest control unit within the caliper of the treated unit, which makes more sense and should yield better matches. Because the old version used a random selection of units, it would be impossible to replicate results without setting a seed prior to calling matchit(). Now, the matched set will be the same every time when a caliper is used. So, even though older results are no longer replicable, the new version improves replicability moving forward and should produce better matches, especially when wider calipers are used.

from matchit.

Thank you so much for the clarification!

from matchit.

Can I ask a follow-up question?

With the new behavior of caliper, does it mean that the matched controls selected when using caliper are always a subset of those selected without using caliper, given all the other parameters are the same (using method = "nearest")?
Since caliper behaviors only helps to remove those "outliers" but for those matched ones they are always the nearest get selected, there's no randomness any more?

Thank you so much!

from matchit.