Rafael Laboissiere | 5 May 21:42 2005
Picon

Re: Statistical analysis of the DPL 2005 election

* Rafael Laboissiere <rafael <at> debian.org> [2005-04-22 12:07]:

> I did a statistical analysis (factor analysis) of the DPL 2005 election
> results and put the conclusions at:
> 
>     http://people.debian.org/~rafael/dpl-vote-2005-analysis/

I added to the paper above some comments by Steve Greenland and Chris
Lawrence.  Here are the relevant paragraphs:

    Factor#2 - the Anthony Towns factor:
        [...]
        In a personal communication, Steve Greenland suggested that AT's
        technical skills were irrelevant and that the reaction for or
        against AT was largely based on these two components: 
            1. AT had proposed temporarily limiting access to the mailling
               lists for people who violated standards of conduct. Steve
	       suspects that this produced a strong reaction of either
	       it's about time or completely out of the question.
            2. As noted above, AT is active at the infrastructure level 
	       of Debian, which gives him a lot of de-facto power over the 
               project. Steve would guess that some people did not think
               that combining this with the office of DPL was a good
               idea.

    Discussion
        [...]
        One might also question whether it is legitimate to use the rank 
	order in the ballots as numerical values for the FA. In a
        private communication, Chris Lawrence argued that it may be better
(Continue reading)

Anthony Towns | 6 May 01:53 2005
Picon

Re: Statistical analysis of the DPL 2005 election

Rafael Laboissiere wrote:
> * Rafael Laboissiere <rafael <at> debian.org> [2005-04-22 12:07]:
>>I did a statistical analysis (factor analysis) of the DPL 2005 election
>>results and put the conclusions at:
>>    http://people.debian.org/~rafael/dpl-vote-2005-analysis/
> I added to the paper above some comments by Steve Greenland and Chris
> Lawrence.  Here are the relevant paragraphs:
>     Factor#2 - the Anthony Towns factor:
>         In a personal communication, Steve Greenland suggested that AT's
>         technical skills were irrelevant [...]

Uh, the point of analyses like these is to get an idea of trends without 
being overly influenced by what your preconceived notions are -- and 
claims of the form "foo was irrelevant" seem particularly inappropriate 
in a statistical context if they aren't directly supported. In any 
event, the conclusions should arise from the statistics, not be added on 
as a post-hoc explanation -- you could as well say "Some have suggested 
AT's platform was irrelevant, and the influence of AT's candidature was 
solely due to the alignment of Mars and Jupiter relative to the moon". 
Maybe that seems reasonable, maybe it doesn't -- but there's nothing in 
the statistics to differentiate that explanation from the ones presented.

Personally, while I would've expected more or less the reaction Steve 
suggests; that doesn't seem to what actually happened: by all the 
measures I could easily do, and while some people certainly voted 
strongly for or against me, I couldn't find any significant difference 
as to how folks voted for me compared to how they voted for Branden; 
except (obviously) that he got a few more votes in his favour.

(It might be interesting to see if the factors still apply when the 
(Continue reading)

Rafael Laboissiere | 9 May 00:53 2005
Picon

Re: Statistical analysis of the DPL 2005 election

* Anthony Towns <aj <at> azure.humbug.org.au> [2005-05-06 09:53]:

> Uh, the point of analyses like these is to get an idea of trends without 
> being overly influenced by what your preconceived notions are -- and 
> claims of the form "foo was irrelevant" seem particularly inappropriate 
> in a statistical context if they aren't directly supported. In any 
> event, the conclusions should arise from the statistics, not be added on 
> as a post-hoc explanation -- you could as well say "Some have suggested 
> AT's platform was irrelevant, and the influence of AT's candidature was 
> solely due to the alignment of Mars and Jupiter relative to the moon". 
> Maybe that seems reasonable, maybe it doesn't -- but there's nothing in 
> the statistics to differentiate that explanation from the ones presented.

Well, the whole "Discussion" section was already subjective and
speculative even before I added Steve's interpretation.  I think I have
made it clear from the beginning.  That said, your comment above is not
wrong.  If you wish, I can add a paragraph with your comments.  Just send
me privately what you wish to be added to the paper.

> Personally, while I would've expected more or less the reaction Steve 
> suggests; that doesn't seem to what actually happened: by all the 
> measures I could easily do, and while some people certainly voted 
> strongly for or against me, I couldn't find any significant difference 
> as to how folks voted for me compared to how they voted for Branden; 
> except (obviously) that he got a few more votes in his favour.

One obvious conclusion from the Factor Analysis is that the election was
polarized around the AT option, meaning that voters tended to rank AT
either above or below all the others (this is factor #2).  OTOH, Branden
and Andreas received similar ranks and they were opposed to Matthew
(Continue reading)

Anthony Towns | 9 May 06:42 2005
Picon

Re: Statistical analysis of the DPL 2005 election

Rafael Laboissiere wrote:
> Well, the whole "Discussion" section was already subjective and
> speculative even before I added Steve's interpretation.  I think I have
> made it clear from the beginning.

Well, subjective and speculative is fine -- but is completely irrelevant 
to the statistics. We've got plenty of places where you and others can 
be subjective and speculative about the results; the whole point of 
doing a statistical analysis is to be evidence and fact driven, surely?

> That said, your comment above is not
> wrong.  If you wish, I can add a paragraph with your comments.  Just send
> me privately what you wish to be added to the paper.

I can understand you including your own speculations -- I mean, you 
wrote the analysis -- but I don't think that's a good idea, and it 
certainly seems a bad idea to include anyone else's.

> One obvious conclusion from the Factor Analysis is that the election was
> polarized around the AT option, meaning that voters tended to rank AT
> either above or below all the others (this is factor #2).

Err. By my count, I was ranked last on 50 ballots, Branden was ranked 
last on 65 ballots. Branden likewise was ranked first on more ballots 
than I was (144 versus 136). If you don't canonicalise the results, the 
only rankings I receive more of the Branden are [3], [4], and [5], 
Branden gets more of each of [1], [2], [6], [7], and [-].

(To get number of ballots ranking each candidate last:

(Continue reading)

Steinar H. Gunderson | 9 May 12:09 2005
Picon

Re: Statistical analysis of the DPL 2005 election

On Mon, May 09, 2005 at 02:42:54PM +1000, Anthony Towns wrote:
> I don't know the statistics; but if "factor" means something similar to 
> "vector basis", then I would've said it meant that "in choosing 
> rankings, people tended to rank AT independently of other candidates".

This is my understanding of it as well.

BTW, does the factor analysis say anything about which factors tended to be
stronger? It looks to me as if all the vectors are normalized.

/* Steinar */
--

-- 
Homepage: http://www.sesse.net/

Rafael Laboissiere | 10 May 09:07 2005
Picon

Re: Statistical analysis of the DPL 2005 election

* Anthony Towns <aj <at> azure.humbug.org.au> [2005-05-09 14:42]:

> Well, subjective and speculative is fine -- but is completely
> irrelevant to the statistics. We've got plenty of places where you and
> others can be subjective and speculative about the results; the whole
> point of doing a statistical analysis is to be evidence and fact
> driven, surely?

Yes.  Perhaps I should change the title of my paper...

> I can understand you including your own speculations -- I mean, you 
> wrote the analysis -- but I don't think that's a good idea, and it 
> certainly seems a bad idea to include anyone else's.

I think the opinions are sound, albeit necessarily speculative. Of
course, this is a matter of personal opinion, but I really do not see a
problem in including the opinion of others in in the paper.

> Err. By my count, I was ranked last on 50 ballots, Branden was ranked 
> last on 65 ballots. Branden likewise was ranked first on more ballots 
> than I was (144 versus 136). If you don't canonicalise the results, the 
> only rankings I receive more of the Branden are [3], [4], and [5], 
> Branden gets more of each of [1], [2], [6], [7], and [-].
> 
> So, afaics there's some misinterpretation going on there: if "AT versus 
> the world" was the meaning of the "factor", "BR versus the world" should 
> have been a slightly stronger one.

These are sensible arguments.  I have to think further on that.

(Continue reading)

Rafael Laboissiere | 24 May 12:05 2005
Picon

Re: Statistical analysis of the DPL 2005 election

[ Sorry for the belated reply.  I am not subscribed to debian-vote and
  noticed your post only today. ]

* Steinar H. Gunderson <sgunderson <at> bigfoot.com> [2005-05-09 12:09]:

> BTW, does the factor analysis say anything about which factors tended
> to be stronger? It looks to me as if all the vectors are normalized.

The vectors are indeed normalized.  The strength of each factor appears
as the proportion of variance explained, as reported in the paper
(Results section).

--

-- 
Rafael


Gmane