Psmatch2, Troubles identifying matched pairs

[Emma Caspersen]

Hi
I am doing Nearest Neighbour propensity score matching and have used the following set of commands.

The identification of matched pairs:

It is not the matching in itself that is the problem, but the problem arises when i have to identify the matched pairs. I have tried the code above, which I have seen suggested on many pages here on statalist. But my problem is that the paircount don't get the value 2 at all. Instead there is values like 0, 1, 14, 27, 40 and 183.

How can this be? And how can I solve this problem?

I hope you can help me solve my problem, as the analysis is a part of my masters thesis.

[Sandra Bloem]

I'm just a student as well (working on my masters thesis as well doing something very similar), so take everything I say with a grain of salt.

Try adding -noreplacement- option to the -psmatch2- command, something like this.

psmatch2 treatment, outcome(y) pscore(propensityscore) common neighbor(1) noreplacement

Without adding this, the control units could be used to form a matched pair with multiple treatment units. For example, let's say we have two treated individuals and two control individuals (table below). Stata will find for every treatment unit the control unit with the closest propensity score. So, for both individual A and B the matched pair would be with individual D.
When using the noreplacement option, what you're telling Stata is that individual C can only be used once to match with a treatment individual, and individual D can only be used once. What is really important before using this is to make sure the observations in your dataset are sorted randomly. I hope this solves your problem!