[Gretl-users] I would rather keep the 31 character name limit than shorten it.
On Wed, 1 May 2019, Sven Schreiber wrote:
> Am 01.05.2019 um 04:53 schrieb Fred Engst:
>
>> 2. With time series data, if the names of two variables that are very
>> long, close to the 31 character limit, and they differ only at the
>> last one or two characters, then a regression of the two with lagged
>> variables will clash with gretl auto created shortened
>> abbreviations.
>
> Certainly a limitation, not sure if it counts as a bug. As the person
> who called for the extension of variable names years ago (from a much
> shorter limit) I see what you mean, but I guess it's OK to ask the user
> to stay below 28 or 25 characters or so.
Yes, I think we should place a limit of (say) 22 or 24 bytes on
user-specified series names. This would leave room for the prefixes
and suffixes that get added automatically in generating logs, squares,
lags and so on.
If we do that I guess we could still accept names of up to 31 bytes in
third-party data files, but should probably issue a warning.
Allin
Hi Allin,
Thanks for all the great responses.
But I’d rather keep the 31 character name than shorten it.
There should be a way to find the maximum length of all variables, and then creating
internal variables or lagged variables based on this information.
I know this will create more work for you. (:
I encountered this problem in my work where I always like to have long names to identify
clearly what the variables are, but found error running regressions with lagged variables.
After some trial and error, I finally isolated the problem as I reported earlier. Now I
know what the problem is, and I’d rather keep the long name and will find a way to work
around the naming limitation with time series. So please don’t limit the variable names to
less then 31.
Fred