On Thu, 6 Dec 2018, Riccardo (Jack) Lucchetti wrote:
On Wed, 5 Dec 2018, Riccardo (Jack) Lucchetti wrote:
> On Wed, 5 Dec 2018, Allin Cottrell wrote:
>> On Wed, 5 Dec 2018, Sven Schreiber wrote:
>>> Am 05.12.2018 um 02:47 schrieb Allin Cottrell:
>>>> Sven called for a new release a while back now. So far as I'm
>>>> we're about ready. Anyone want to call for a delay? (If so, please
>>>> explain, briefly.)
>>> Depends on what you call a delay. Personally I guess I have to update the
>>> German translation which should be possible by Friday.
>>> Then there's the small package menu attachment bug I mentioned off-list
>>> yesterday. And the new thing from Ioannis today.
>> OK, so aim for a release next week?
> Fine by me!
Uhm, hold on a second. Is this intended?
setobs 5 1:1 --stacked-cross-section
series u = uniform() - 0.5
series x = u
x = 0.5 * x(-1) + u
print u x -o
? print u x -o
1:1 0.1531123 0.1531123
1:2 -0.3014429 -0.2248868
1:3 0.1214166 0.008973
No, not at all intended. That bug is now fixed in git and snapshots.
Just for the record, the bug Jack exposed is that when an
autoregressive "genr" expression, such as
y = y(-1) + u
was evaluated in a panel dataset, the resulting series was all
missing values apart from the first individual in the panel. This
was a regression associated with the change in gretl's internal
definition of "NA" or missing that took place in git in August of
this year. But that change was reverted for the 2018c release, since
it hadn't been tested fully enough, so the problem is not present in