Re: nm.debian.org statistics buggy ?
On Fri, Aug 16, 2002 at 02:59:17PM +0200, Helios de Creisquer wrote:
> Without any crystal ball, I am able to say that an applicant waiting
> since 3 months at this stage will have a total time of waiting superior
> to 3 months for this stage when his complete application will be done.
> AFAIK that's not called divination, but simple logic. Am I wrong ?
Yes you are, or not thinking the whole problem through. You've just
worked out the maximum, or rather the minimum, or...
What is it exactly?
You've worked out the minimal value that the maximum field can be,
assuming that this applicant is processed today, of
course its wrong tomorrow.
Now, what about the averages? Do we consider the person who's AM
recommended him today in the set?
If we had four people in the queue, one recommended today, one
recommended 4 months ago and two recommended 2 months ago. The averages
are all 2 months. Move them around a bit, then skew the statistics
where there are a small number of outliers, a lot in the middle and
a fair few at the smaller end of the scale and you see it all becomes
a mess quickly.
What we currently have is the least-worst set of statistics, certainly
not the perfect set.
> So why not include not_yet_fully_processed_applicants in the statistics
> page ?
Because it is meaningless, the long waiting people are there due to
reasons (for those that are going to ask, I have no idea - timezone
problems mean I rarely talk to James). I'd say a large majority
of people do not fit the category.
Craig Small VK2XLZ GnuPG:1C1B D893 1418 2AF4 45EE 95CB C76C E5AC 12CA DFA5
Eye-Net Consulting http://www.eye-net.com.au/ <email@example.com>
MIEEE <firstname.lastname@example.org> Debian developer <email@example.com>