teffects nmatch

[Jack Ryan]

Hey Statalist,

I need help using the teffects nnmatch - nearest-neighbor matching function in Stata. I try to investigate the effect of an event using this function. There is a group that have received a treatment (Treatment Group) and a group that has not received a treatment (Control Group). I use this function to prove this effect by calculating the Average Treatment Effect on Treatment. Now I have the problem that I can calculate the ATT by using the function effects nmatch, but I don't know which treatment part was matched with which control part. I want to look at the details of the matched pairs and compare them. It would be ideally, if e.g. Bank A is matched with Bank C and E (in the case of nneighbor(2)), to export all the information into new variables. So that I could say something about the total_assets of Bank A in comparison to Bank C and E.

My function looks like that:

teffects nnmatch (d_ct1_capital total_assets) (treatment) if first_obs==1, atet nneighbor(2) vce(iid) metric(maha) biasadj(total_assets)

I read in a different post, that with psmatch you may use the function keep() to extract the needed information about the matches

But I don't find a solution for the teffects nmatch case

Thank you for your help

[Dimitriy V. Masterov]

Cross-posted on SO.

[Nick Cox]

The Stack Overflow post was put on hold. It's so off-topic there that no-one felt to obliged to explain why in detail.

Dimitriy V. Masterov was, I guess, alluding to our policy on cross-posting, which is explicit in the FAQ Advice all members are asked to read in posting: You are asked to tell us about cross-posting.

This hasn't got a direct answer yet, for which there are various hypotheses:

1. Time of year. In many countries, people's energies are directed elsewhere.

2. You are being punished silently for an implausible name. If "Jack Ryan" really is your real name, then commiserations, because you are likely to be the subject of weak jokes made frequently. But many people will be guessing that it's a fake identifier, whereas (FAQ Advice again) you are asked to use a full real name here.

3. "I need help" is vacuously true of any question here. It's not a good opener for a deeper reason. Statalist is not a help line; it's a forum in which posters are asked to phrase questions so that they can be answered: that's to their advantage too, as there is no point in posting otherwise. The FAQ Advice makes several suggestions which you didn't follow.

4. It's hard for experienced Stata users to read the question because it ignores Stata terminology in favour of something else. Naturally we recognise that many people come here with more knowledge of some other software than of Stata, but the point remains.

Below I post your question again, but using Stata terminology and edited to remove several apparent typos and to indicate better style, gaps and puzzles. I appreciate that English is quite possibly not your first language. Using a different font for Stata code isn't even recommended explicitly, but it helps to make your question more readable.

!!! revised question begins

I am using the teffects nnmatch nearest-neighbor matching command. I am investigating the effect of an event. I have a Treatment Group and a Control Group. I am using this command to prove this effect by calculating the Average Treatment Effect.

I can calculate the ATT [?ATE] by using teffects nnmatch, but I don't know which treatment part was matched with which control part. I want to look at the details of the matched pairs and compare them. It would be ideal if e.g. Bank A is matched with Bank C and E (in the case of nneighbor(2)), to export all the information into new variables. So that I could say something about the total_assets of Bank A in comparison with Banks C and E.

My code looks like

teffects nnmatch (d_ct1_capital total_assets) (treatment) if first_obs==1, atet nneighbor(2) vce(iid) metric(maha) biasadj(total_assets)

I read in a different post [???cross-reference], that with psmatch you may use the option keep() to extract the needed information about the matches

But I can't find a solution for the teffects nnmatch case.

!!! revised question ends

I stop there. I don't ever use this stuff and I have no idea whether your question can be answered and/or what other information may be needed. But I have given your question a better chance. Points 2, 3, 4 remain for your future posts. More generally, when we ask "Please read the FAQ Advice" we do really mean that.

[Jack Ryan]

Good evening Nick Cox,

I'm very sorry I didn't follow the rules. It was my first post and is therefore apparently full of mistakes and inaccuracies. I apologize for that and it will not happen again in the future. Thank you for clarifying it.

Best,

Jack