On Sun, Feb 17, 2013 at 10:41 PM, Amgine <amgine(a)wikimedians.ca> wrote:
On 17/02/13 09:18 PM, James Salsman wrote:
I've been working as a professional statistician for most of the past
two decades, in applied fields where problems result in immediate
failures of various algorithms. These opinion questions aren't
sophisticated. People will answer when they are asked. What do you
think makes sampling so difficult, in the case, for instance, of how
often advocacy actions should be proposed?
Well, as a simple issue of weighting, should geographic weighting be applied
based on size of a given language speaking population? e.g. Should the
sample of English speaking respondents from India be relevant to the
region's weight in English as a language, it's population as a portion of
humanity, its estimated online population, or some other global-relative
weighting?
No, randomly selecting editors with email registered from those who
have edited on any project a certain number of times in the past
month, say five times to use one of the Foundation's definitions for
active editors, and asking them their opinion in their primary
language as determined by the project they have edited most frequently
over that period, will result in an answer which reflects the
geographic distribution of the editor community.
Or how about projects: are all project populations to
be the same size?
No, but their number of active editors reflects the proportion in
which the Foundation should reasonably weight their editors' opinions
of, in this instance, how often advocacy actions should take place.
I'd be very interested in what field you are a
professional statistician; my
area is health sciences, particularly public health.
My most recent professional publication is
https://docs.google.com/file/d/0B73LgocyHQnfS0g5ZEw1aFNKT2s/edit but
the bulk of my income over the past year has been from work done for
commodities futures traders.