Re: Ian et al: Poll conclusions

Arne Babenhauserheide <arne_bab-S0/[email protected]> Thu, 08 Dec 2016 19:52:31 +0100
Newsgroups	gmane.network.freenet.devel
Message-ID	<[email protected]>
[email protected] writes:

> On Tuesday, December 06, 2016 06:49:04 PM Arne Babenhauserheide wrote:
>> It’s been over a week now, could you please post the data with or
>> without the evaluation?
>
> I offered something like that at both the beginning and the end of my reply to 
> Ian the last week and got no "Yes, do it" from him.
>
> Nevertheless if he doesn't reply to the mail within a day, I will assume he 
> wants that and this week allocate a day for finishing the evaluation as I 
> consider it reasonable and publish it then.

That’s good. I don’t think we’re doing Ian a service when we make him a
bottleneck for things which block other tasks.

> Meanwhile, please refrain from your attempts to make me throw away days of 
> work of manually putting the stuff into spreadsheets, reviewing and coloring 
> it, etc., just so you can replace it with scripts which will be blind for all 
> the things a manual evaluation is not blind for.
> You can run your scripts on the raw data afterwards, but I really don't feel 
> like throwing away days of my work in favor of simplistic scripts.

I’m sorry that you got the notion that the scripts are simplistic. The
implemented approach to derive a Condorcet method and use different
evaluations to provide a robust top 5 from the data isn’t
simplistic. It’s the best result I can get from the data without adding
subjective judgement about the votes people donated to the project.

> I *am* thankful for your efforts, but I doubt they'd spot the borderline-
> fraudulent data which a human can clearly see.

I also looked at all the results manually, but I think that using
clearly defined criteria (as in the scripts) is a better method than
relying on subjective human marking, because what looks fraudulent to
one might just be a strong difference in priorities between different
groups in our community.

I don’t mind doing an evaluation which includes manual human judgement
and one which does not, so we can compare the results.

> Please be aware that the spreadsheets *do* contain some of your suggestions 
> already such as median, standard deviation, etc., so your work is not 
> neglected either :)

That sounds good.

Best wishes,
Arne
-- 
Unpolitisch sein
heißt politisch sein
ohne es zu merken

_______________________________________________
Devl mailing list
[email protected]
https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
signature.asc (application/pgp-signature, 800 B)