-
Notifications
You must be signed in to change notification settings - Fork 82
Open
Labels
minor bugBug unlikely to lead to incorrect analysisBug unlikely to lead to incorrect analysis
Description
If there are too many unique values in a string/categorical variable, levelsof breaks with an error message of "cannot compute". I have just run into this with a variable that had 700k+ unique values.
It now runs with the workaround of replacing the following lines
ietoolkit/src/ado_files/iesave.ado
Lines 599 to 603 in fa1146e
| * Number of levels and complete observations | |
| qui levelsof `var' | |
| local varlevels = r(r) | |
| local varcomplete = r(N) | |
with
* Number of levels
preserve
keep `var'
duplicates drop
count
local varlevels = r(r)
restore
* Number of complete observations
qui count if !missing(`var')
local varcomplete = r(N)
There may be a more elegant approach, though. If no one can think of one, I can open a PR with this one.
Reactions are currently unavailable
Metadata
Metadata
Assignees
Labels
minor bugBug unlikely to lead to incorrect analysisBug unlikely to lead to incorrect analysis