Matching on multiple variables

[Andrey Dolinko]

Hello,
1st time poster here, so please let me know if you need additional details to help.

I am trying to match cases to controls in a 1:2 ratio, ideally on three variables in a specific order:
1. Age +/- 0.5 years
2. lab value (with specific upper and lower calipers)
3. BMI +/- 3 kg/m2

I have tried to use -rangejoin- but seem to have two limitations:
First, I'm not sure if I can match by an age range? I have tried doing it be exact age for now.
Second, I cannot figure out how to match by the last variable.

I would appreciate any help!
Thank you,
Andrey

[George Ford]

You should probably include all the data in a single dta, mark the treated, and then use a matching algorithm (entropy matching, CEM, propensity, etc).

[Andrey Dolinko]

George - Thank you for the suggestion.
Is there one matching algorithm that you would recommend over another?
At this moment I am not trying to identify a specific statistic (e.g. propensity score), but truly just trying to see if I have the appropriate patient data collected or if I need to collect more.
That's where I found -rangejoin- helpful in helping me identify which case matched to which control.
Is there a way to do this with one of the matching algorithms? How do I set the calipers appropriately for each of the variables?

[George Ford]

There are many, and kmatch does most of them. Propensity score is probably the most common and a safe bet. I've found entropy matching to be fast and very good at the task, and kmatch is much faster at CEM than is cem.