On Mon, 14 Jan 2019, Sven Schreiber wrote:
Am 14.01.2019 um 15:46 schrieb Allin Cottrell:
> On Mon, 14 Jan 2019, Allin Cottrell wrote:
>
>> On Mon, 14 Jan 2019, Javier Sansa wrote:
>>> d_x = x- x(-1)
>>>
>>> is used, the result is a differenced series that fills the whole sample.
>>
>> True. Using the diff() function produces the same result as your formula
>> above, but the diff command is not doing the right thing in respect of a
>> lengthened sample range. We'll figure out what the problem is.
>
> It's now fixed in git. Snapshots will follow before long. This was subtle
> fallout from the internal redefinition of "NA" since 2018c.
>
I'm glad that a new version wasn't already released before this fix, but the
two recent bugs together are really problematic IMHO, because they affect
general usage and no special niche case. So I'm now quite firmly in favor of
a new release.
I agree. If anyone wants to update translations, please go ahead. I
will aim to put out 2019a by the end of this week.
Allin