lubuntu-desktop team mailing list archive

[Jonathan Marsden <jmarsden@xxxxxxxxxxx>]

On 12/17/2011 04:36 PM, Philip wrote:

> Would it be possible to include something like a voluntary survey
> with a round of updates? That would be better than raw numbers, and
> the users who care enough to fill it out would probably provide the
> best data.


Would this definitely be "better than raw numbers"?  How do you know
this?  That's a *very* bold statement -- do you have data to back it up?

It could accidentally be a very biased sample; "users who care enough"
might all be significantly more experienced and techie than "users who
do not care enough", for example, and so exhibit rather different usage
patterns.  And now, if we were to trust the survey data, we might
accidentally skew our decisions based on this biased sample, and lose
our less-techie users... not what we want.  Or, maybe, our experienced
users all have lots of work to do and lots of spam email, and will
reflexively reject all survey requests, but newcomers are so
enthusiastic that they will fill them in; in that case you could get the
opposite bias.

It seems to me that you cannot be sure your survey data would be
"better" unless you have done all kinds of work (perhaps field trials,
and appropriate statistical analysis of the results) to demonstrate
that.  Have you in fact done that work?

As you must be aware, since you will already have researched this field
in some depth before proposing something new, the existing
popularity-contest package is a well established attempt to gather this
kind of data, from all those who enable it, based on what people
install, rather than on what they run.  Debian automatically graphs the
data it provides, all the way back to at least 2004, for example:

  http://qa.debian.org/popcon.php?package=bash

There is Ubuntu data at http://popcon.ubuntu.com too.  I do not know
whether popcon in its current form can distinguish between data from
Lubuntu and data from Ubuntu; I suspect not.

Will whatever new survey you are proposing clearly do better than this?
 Would enhancing popularity-contest and its associated infrastructure so
it *does* distinguish between the various flavours of Ubuntu be as
useful, or more useful, than your proposal?  How do current numbers of
users (as reported by popcon) installing lubuntu-desktop or lxde-common
compare to those installing ubuntu-desktop, for example?

Overall, I recommend using existing tools and data over trying to build
a new survey mechanism from scratch.

Jonathan