On Mon, May 3, 2010 at 10:45 PM, Dave Airlie wrote:
On Mon, 2010-05-03 at 22:37 -0400, Orcan Ogetbil wrote:
> On Mon, May 3, 2010 at 10:26 PM, Chris Adams wrote:
> > Once upon a time, Orcan Ogetbil said:
> >> The statistic talks. It doesn't only talk. It yells. Ignoring this
> >> test statistic in favor of the large pool of imaginary users, who
> >> supposedly think in the complete other direction, is not only
> >> non-scientific, stupid., but also self-conflicting.
> >
> > You claim you have taken grad-level statistics, but you don't appear to
> > understand how to select a sample. A self-selected sample on a web
> > forum (with no basis to show that forum members are representative of
> > the Fedora user base, much less that poll responders represent even the
> > forum users) does not lead to valid statistical results. It doesn't
> > matter how big the sample size is if the samples are not properly
> > selected.
> >
> >
>
> No, I exactly know how to select a sample. The forum based sample is
> "the" perfect sample. It is the users who talk. It is not the
> imaginary users, who you claim to exist. I do not care about your
> imaginary users. I do care about those who talk. Hence the statistic
> is significant.
You aren't a very good statistician if you care. thats an implicit bias.
... which is far better than an imaginary bias.
Orcan
PS: No, I am not a statistician. But I know about the science of
statistics to a good degree.