[Foundation-l] Attribution survey, first results

[phoebe ayers]

On Wed, Mar 4, 2009 at 11:08 AM, Mike Godwin <mnemonic at gmail.com> wrote:
> Phoebe writes:
>
> This is a very small, self-selected sample; there would be
>> no harm or cost associated with turning it on for a much larger
>> percentage (or all) of logged-in users on the top-ten languages, not
>> just English or German alone, which both have peculiarities associated
>> with being the largest Wikipedia communities.
>
>
> Is there a version of the survey that does *not* entail a self-selected
> sample?  The methodologist in me wants to know, because it seems to me that
> selection bias is inherent in any survey of this sort. (What's more, it
> seems fairly predictable in which direction that bias would skew results.)
>
>
> --Mike

This is true -- though I'm no statistician and I'd love to hear from
those who are (Robert?) I have wrestled with this question before and
I'm not sure there's any way to get a non-self-selected survey about
anything on the projects due to anonymity concerns. However, I'm
mostly concerned with:

* getting responses from more language communities
* it seems there may be a difference between those who see the notice
popup on their watchlists (the 5%) and those who chose to go to the
survey to take it (people like me, who really care about this issue).
For instance, we ran an item about the survey in yesterday's
English-Wikipedia signpost; enough people read this that could cause a
spike in responses. I'm not sure if this is a valid methodological
concern or not -- or perhaps we are mostly interested in getting
responses from people who really care? I'm not sure.
* the short time frame arguably leaves out the class of editors who
only log in occasionally; their responses may be different from the
editing-every-day crowd due to a qualitative difference in
participation. (Or not! Who knows).

Also, some clarity in what each of the options means would be good;
the question about participation in foundation activities in
particular seemed a bit vague. More to the point, while I agree it's
interesting to know what infrequent-editors versus heavy-editors think
about the question, how is participation level and thoughts on
attribution going to be correlated? Do one's thoughts matter less if
one is an infrequent editor? Etc.

One quick way to get some expert feedback about all these questions
would be to submit the survey design to wiki-research-l, where the
researchers hang out.

-- phoebe