5
$\begingroup$

When matching based on the Mahalanobis distance (MD), are there guidelines for selecting the caliper. For example, if the propensity score is used as the distance metric, literature supports starting with 0.2 standard deviations of the logit. Does something similar exist for the MD?

$\endgroup$
2
  • $\begingroup$ Could you maybe provide links to the literature you are referring to? $\endgroup$ Dec 2 '20 at 13:23
  • $\begingroup$ Did you find something satisfying since then ? $\endgroup$
    – keepAlive
    Feb 14 at 8:52
3
$\begingroup$

As with all matching, pick the caliper that yields the best balance after matching. Also consider the number and range of the remaining treated units. The correct caliper will depend on the characteristics of your data set. I know this doesn't really answer your question, but you should know that the ubiquitous 0.2 standard deviations of the logit of the propensity score is arbitrary.

$\endgroup$
1
  • 1
    $\begingroup$ Thanks. Just to clarify. I meant the 0.2 SD as a starting point. There is literature supporting this choice so hopefully it is not completely arbitrary. $\endgroup$
    – julieth
    Jul 28 '17 at 23:52
2
$\begingroup$

Despite Gary King pointing out the issues with Propensity score matching a few years ago, propensity score matching seems to still be the most commonly used distance metric for matching, and all calipers I have been able to find is based on it.

This paper was the closest thing I could find to answering your question:

https://www.lexjansen.com/pharmasug/2006/PublicHealthResearch/PR05.pdf

If you have found a better answer to your own question in the meanwhile, please do inform me about it.

$\endgroup$
1
  • 1
    $\begingroup$ Gary King seems to be strongly against the use of propensity scores for matching under any circumstances and instead proposes Mahalanobis based matching as a 'better' alternative. Can anyone provide reasoning for the proposed method of combining propensity score and Mahalanobis matching? Under which circumstances would this method be 'better' than a purely Mahalanobis based method? $\endgroup$ Dec 2 '20 at 13:35

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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