W3C home > Mailing lists > Public > public-html@w3.org > February 2009

Re: summary="" in HTML5 ISSUE-32

From: Robert J Burns <rob@robburns.com>
Date: Thu, 26 Feb 2009 21:44:56 -0500
Cc: Ian Hickson <ian@hixie.ch>, Steven Faulkner <faulkner.steve@gmail.com>, HTMLWG <public-html@w3.org>, W3C WAI Protocols & Formats <w3c-wai-pf@w3.org>
Message-Id: <A23548C8-8186-44DA-B6B7-49C0120C7FED@robburns.com>
To: Maciej Stachowiak <mjs@apple.com>
Hi Maciej,

On Feb 26, 2009, at 9:31 PM, Maciej Stachowiak wrote:
>
> On Feb 26, 2009, at 6:13 PM, Robert J Burns wrote:
>
>> HI Maciej,
>>
>> On Feb 26, 2009, at 8:48 PM, Maciej Stachowiak wrote:
>>
>>>
>>> On Feb 26, 2009, at 5:29 PM, Robert J Burns wrote:
>>>
>>>>
>>>>> If you have more objective data, then by all means, present it.
>>>>
>>>> The scientific method does not say that if we can find suitable  
>>>> data than we can draw conclusions from it, but if we cannot then  
>>>> we should simply allow one person in a position of leadership to  
>>>> make wild speculation about what data might possibly exist if we  
>>>> had the resources to acquire it.
>>>
>>> The scientific method says that to dispute a theory (e.g. "summary  
>>> values are usually poor"), one provides contrary evidence. For  
>>> example, showing actual selection bias in Philip's study, or doing  
>>> a study that shows different results, would be examples of the  
>>> scientific method in action. Contrary evidence is what we use to  
>>> reject a hypothesis.
>>
>> Its what the scientific community uses to reject hypothesis. You  
>> use PR style spin.
>
> That seems like an unwarranted remark. I don't believe I have used  
> "PR style spin" to reject a hypothesis. I'm not sure what that would  
> even entail.

See below.

>
>
>>
>>> On the other hand, spreading fear, uncertainty and doubt about the  
>>> quality of Philip's study is not the scientific method. It is not  
>>> science to say the study is bad without pointing out specific  
>>> problems or showing contradictory results.
>>
>> I'm not spreading fear, uncertainty and doubt about the quality of  
>> Philip's study. You're making wild claims about what the study says  
>> in suggesting that I am. Philip himself disclosed the bias of the  
>> sample. I don't even care about that so much. Let's use it as  
>> anecdotal evidence which can still be useful. But let us use  
>> anecdotal evidence for what it is useful for and not claim it can  
>> show us what it cannot. For example it cannot show us what  
>> percentage of pages use summary='' for layout tables or misuse is  
>> it for non-layout tables.
>
> I don't recall endorsing Philip's study or making any conclusions  
> about what it claims.

You claimed that I was spreading feat uncertainty and doubt simply  
because I said Philips study had bias in a way that didn't serve the  
needs of the WG

> Philip says he did not use a high quality method to pick a random  
> sample.

I don't think Philip said he did not use a high quality method. Now  
you're impugning Philip's methods even though I did not.

> But that does not in itself prove a sample bias or selection bias,  
> in particular it does not show the results are non-representative of  
> how summary="" is used.

Yes it is a biased sampled in such a way that it does not really serve  
the needs of the WG. So as I said lets use it for anecdotal evidence  
which can still be useful. However, so can many other deliberation and  
discussion techniques that have been dismissed by you, Ian, and others.

> It is well known in statistics that fairly small samples can lead to  
> statistically significant conclusions about the population as a whole.
>
> If you'd like to show the results are wrong, you have to show some  
> evidence, not just cast vague aspersions on the methodology.

That's it right there. You just took a study that does not follow  
statistical methods to lead to an unbiased sample draw (as Philip  
acknowledges), claimed it did, and then told me I must find a way to  
prove that it didn't. That is PR style spin.

In what way could I "prove" that the Philip's sample had a bias? How  
can I prove that we're developing the next version of HTML for use on  
the web and off the web in places other than those indexed on  
dmoz.org? There's no way to respond to that with scientific evidence  
because we are no longer in the realm of science. We are no simply  
using PR-style spin. If instead you can also acknowledge that the  
sample was drawn from a population biased toward a population which  
was not our only concern, then we could proceed to talk about it in  
more reasoned terms: in particular for its anecdotal qualities.

Take care,
Rob
Received on Friday, 27 February 2009 02:45:42 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 9 May 2012 00:16:32 GMT