Re: Vote for Linux: stats

Alain KNAFF (Alain.Knaff@imag.fr)
Tue, 16 Jul 1996 11:18:50 +0200


>On Sun, 14 Jul 1996, Alan Cox wrote:
>
>> > the real score is *drumroll*: 2921 Yes and 229 No.
>> > Yes, that's 92% against 8% :-)
>>
>> However it doesnt alter the fact people have been making multiple votes and
>> the like (both for and against). Don't count the vote as of any big value.
>
>If you look now they show it again as 53% against 47%, now that the voting
>is over.

This display was actually an error, else why would they have stated
that a majority voted yes, and showed a graph where the majority voted
no? I e-mailed them about it, and they acknowleged the error, saying
they would correct it later that day. However, at first, they didn't
change the graph at all.
Then, this morning, there was indeed a change: Now all reference to
that poll has disappeared, and now it shows the results of another
poll: "Would you use a future version of Notes as a Web server?".
Maybe it's the poll from two weeks ago?. I mailed them again about it,
now let's wait and see what'll happen next...

>They probably sorted out all the double, tripple and quad votes....

The quickpoll.dat file doesn't contain enough information to sort out
these alleged double, tripple and quad votes. Most importantly, it
lacks the hostname where these originated from. Just looking at the
string of yes and no votes is not enough. If you have ten times more
yes votes than no votes, there is a 100% certainty that there is at
least one string of 9 yes votes in a row, and there is a very high
probability that there are many of these long "yes" strings. It's
basic number theory and statistics, just think about it.

Alain