CSDID Characteristic Contents Too Long

[Trevor Andrew]

I am using the csdid command

csdid outcome, ivar(place) time(time) gvar(gvar) notyet

and get the following error:

Characteristic contents too long. The maximum value of the contents is 67,784.

Has anyone run into this problem? The observation level is month-place and I have around 110k observations.

[FernandoRios]

Can you "tab" gvar time. That, i guess, will be the biggest issue

[Trevor Andrew]

It says too many values

[FernandoRios]

That may be your answer.
Unfortunately, without that info, I cant help with any trouble shooting.

[Trevor Andrew]

I have 180 distinct time values and 57 distinct gvar values. Is there a problem for the command if you have too many time or gvar values?

[FernandoRios]

It could be because of how your Gvars are stored. I usually suggest to use a simple conversion (say _10 for first month in data, 11 for the next so on so forht, and modify gvar and tvar in both)
however, 57*180 also means 10K variables.
That could also be biting here.

[Trevor Andrew]

I followed https://asjadnaqvi.github.io/DiD/docs/code/06_05_csdid/ and did the exact same set up. time is 1,..,180. So essentially you are saying use time as _10,11,12,... gvar would be 0 if not treated and if e.g., treated in second month 11 for that unit? Just making sure since I don't understand what you mean by modify gvar and tvar in both?

[FernandoRios]

Well my point is
csdid uses the timing do define pre and post periods and zero for never treated
the issue of a period 1 could be that the pre period is now 0 which is a problem
min any case your issue may be more of memory and long names but can’t really say without access to the data
and of course limited time trying to look into your data

[Trevor Andrew]

Thanks. I'll try to define time starting with 10. I tried running the same command on a restricted number of years and it worked. So maybe it is the memory issue. By long names you mean variable names?

[FernandoRios]

ok now i understand better
I try to save information on all the variables created because those are used internally.
The way to store them is via locals or chars. But because you have very large number of cases, the CHAR that keeps all that data exceeds tStata limits, which is why it is stopping
F