KDE Perceived Quality Survey (PESQUS)

[Benjamin Rossen]

On Tuesday 18 January 2005 16:29, Chad Robinson wrote:
> Benjamin Rossen wrote:
> 
> An interesting point, but did you really think I intended the e-mail 
> addresses  
> submitted to be published? I've never taken, let alone seen, a survey that 
> disclosed the identities of participants, but many collect e-mail addresses. 
> At the risk of sounding too sarcastic, ahem:
> 
>    "unless you mean that we should see the results at the end"
> 
> Yes, that's what I meant when I said:
> 
>    "it's important that survey participants be party to the results"
> 
My apologies. I did not intend to put you down in any way. If it looks like 
that then I am at fault for phrasing my comments poorly. Nor did I think you 
wanted to see other people's e-mail addresses. 

One sometimes see these kinds of surveys showing a running score of the 
aggregates taken so far. You often see that sort of think of television quizz 
shows. I thought you might mean something like that; which seemed implied by 
your use of the continuous tense; "... what the others are doing ..." 

Looking at running scores tends to produce poorer outcomes. If I run a 
competition for people to guess how many jelly beans there are in a jar, and 
a statistically large number of people participate, the average shall be very 
close to the real number, almost always better than the best individual guess 
- and as the number of people guessing gets larger the difference usually 
gets larger unless by chance someone guesses the exact number. Even when this 
happens, we can still say that the distribution of averages and the 
distribution of best guesses show the averages to be better than the best 
guesses; both distributions having their mean on the true number but the best 
guesses having a larger standard deviation. 

The over estimates and the underestimates shall cancel each other out. This is 
a surprising discovery, and is very robust. It has been replicated many many 
times. 

On the other hand, if you put up a running average of the prior guesses, the 
end result shall not be accurate, and the best guesses shall be better that 
the average of all guesses. If the initial few people started out guessing 
too high, the average shall be too high. That too is a robust finding. 

This principle applies in many areas of mass psychology, including filling out 
surveys. 

So, thank you for your contribution to the discussion. I agree with everything 
you said, and if I misunderstood your intended meaning, then please accept 
this apology. 

Benjamin Rossen 
___________________________________________________
This message is from the kde mailing list.
Account management:  https://mail.kde.org/mailman/listinfo/kde.
Archives: http://lists.kde.org/.
More info: http://www.kde.org/faq.html.