Closed sergiocorreia closed 9 years ago
When running the command with a large dataset, I got a hard-to-replicate bug
. sumup contribuyente , by(cat_contribuyente ) Obs. nos. out of range r(198);
After some tracing, it seems this line is the culprit:
bys `touse' `by' : gen `bylength' = _N
If the dataset is too large, you have rounding errors and you may end up doing with values outside the range of the dataset.
If you set the type to long (or double), the issue would be fixed
Also, note that even if double would work, in practice we rarely have datasets above 2BN obs, so a line like this may be useful:
local type = cond(c(N)>c(maxlong), "double", "long") bys `touse' `by' : gen `type' `bylength' = _N
Best, S
Very helpful, thanks.
matthieugomez
commented
9 years ago matthieugomez
commented
9 years ago
When running the command with a large dataset, I got a hard-to-replicate bug
After some tracing, it seems this line is the culprit:
If the dataset is too large, you have rounding errors and you may end up doing with values outside the range of the dataset.
If you set the type to long (or double), the issue would be fixed
Also, note that even if double would work, in practice we rarely have datasets above 2BN obs, so a line like this may be useful:
Best, S