invalid numlist has too many elements with synth

[Rey Juliette]

Dear stata users, I'm trying to run synthetic control group on stata 11.2 to measure the effect of public aids on the recipients. In order to do that, I download the synth package created by Abadie, Diamond and Hainmueller from SSC. My stata command is then synth ca age effectifestimé taille investissement surcapacitéessai , trunit(100015) trperiod(2005) However, I get the following error message, that I don't understand: uno() invalid -- invalid numlist has too many elements Is there any idea of what should I do? Thanks by advance, Juliette

[Nick Cox]

http://stackoverflow.com/questions/3...imits-suggests

I don't know more (in fact I know less) than is discussed there. I've not used this program.

[Rey Juliette]

Thank you for the answer. It seems that the solution is to modify the ado file of synth, but it seems quite difficult to me..

[Sammara Cavalcanti]

Hi, Rey Juliette, did you find the solution to this problem (uno() invalid -- invalid numlist has too many elements)? I have the same problem with "synth" command.

[Alison Levin-Rector]

Does anyone have a solution to this yet?

[Andre Reigado]

I am having the same problem right now. I have already tried to diminish the number of units to less than 2500 and it still has not worked. Even with less than 2000 it does not run.

Does anyone has an idea ?

[Felipe Lozano]

Hi, I just had the same error, but I realized that even if I limited the numbers it wouldn't work until the panel was balanced and not only at the observation level but for the outcomes and dependent variables. Have you tried that??

[Filippo Santi]

I think I am a bit late to be useful to any of you, but the Felipe got it right. You need all variables - outcomes and controls - to be non-missing over the entire period of analysis.

[Felipe Lozano]

Years later I found this error again even after balancing the panel. The suggestion from the link above of altering the numlist limit from the .ado file does not seem easy. So far I have gotten around this by subsampling on the number of donors to meet the limit, at the end the procedure selects only a handful or a dozen. In my case, I want to keep most of the donors that are logically in the proximity (geographically or similar population or other characteristics) of the treated unit. I subsample those first and I fill the limit (1600) with additional units selected randomly.

[Jared Greathouse]

Tell me how many pre-intervention periods you have, please. Felipe Lozano

EDIT: I'm unclear why the number of donors matters here. You're allowed to have more than (in my experience) 300 donors. I do not recommend this (unless you're using my SCM command that isn't public yet), but the number of donors can't be why this error is happening unless you've seen something i haven't