[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Anything wrong with med-bio?

Regarding popcon stats, I think it would be highly useful to be able to
know the following three numbers.  I am not sure which of these three
are available at the moment, but if all three are available then we can
probably due some interesting nonclassical statistics using popcon.
We would need:

1) The total number of machines, M, using popcon.
2) A count of machines that have package i installed F(i) for all packages i
3) A count of machines that have both package i and j installed F(i,j)
for all packages i, j

Using these three statistics, we can compute something that can relate
the typical use of one package to another and put packages in larger groups
for context.  This is the type of analysis that libcomplearn and libqsearch
enable [1].

Best regards,


On Nov 20, 2007 10:22 PM, Andreas Tille <tillea@rki.de> wrote:
> yn Wed, 21 Nov 2007, Charles Plessy wrote:
> > To my surprise, boxshade, which had close to 100 installations recently,
> > dropped strongly.
> Well, sometimes popcon results are hard to interpret.  It might be that
> there was some kind of cluster installed that added say 64 instances of
> installed boxshade (for whatever reason) and got it deinstalled later on.
> Well, I admit that looking at the graph does not look like a cluster
> effect because in this case you would observe a steeper slope.
> > After investigation and comparison with packages like
> > perlprimer and emboss, which are still rocketting,
> It might be that those packages are installed without using med-bio
> which is perfectly possible.
> > I found out that the
> > decrease affected a subset of our packages, that have in commont to be
> > installed by the med-bio metapackage. Indeed, med-bio decreased as well.
> Perhaps we should try to find some correlation between given Debian-Med
> talks and installations of med-bio.  I see two peaks in 2007 and I had
> two chances to talk about Debian-Med
>     http://people.debian.org/~tille/talks/200705_ltb/index_en.html
>     http://people.debian.org/~tille/talks/200706_debconf7_med/index_en.html
> It might be that several visitors of the talks tested, played around
> and deinstalled med-bio.
> > Although there is no indication that there could be a problem with
> > med-bio, I just wanted to let you know this fact in case you noticed
> > something particular. I have the feeling that many users do not report
> > bugs, and that the decreases can indicate problems.
> Well, I really like the idea of observing popcon stats to draw some
> conclusions about the usefullness of our packages.  On the other hand
> we are doing here statistics with quite small numbers.  I would start
> wondering about problems if the number of installations would have
> crossed the borderline of 100 and than dropped below 50.  We should
> know from the people reading the list whether this is some effect
> caused by a cluster (as guessed above) or whether it is something else.
> > For insance, if you
> > look at clustalx, the two decrease periods coincide with buildd gizmo
> > and segfault bug.
> >
> > http://people.debian.org/~igloo/popcon-graphs/index.php?packages=clustalx
> >
> > just in case,
> This sounds like a reasonable interpretation but I'm a little bit reluctant
> to make a science of popcon results because there might be varios different
> influences we just can not know and the absolute numbers we are talking about
> are quite low.
> Kind regards and thanks for sharing your observation
>         Andreas.
> --
> http://fam-tille.de
> --
> To UNSUBSCRIBE, email to debian-med-REQUEST@lists.debian.org
> with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org

"We can try to do it by breaking free of the mental prison of
separation and exclusion and see the world in its interconnectedness
and non-separability, allowing new alternatives to emerge." -- after
Vandana Shiva

Reply to: