W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > October to December 2005

Re: validity.htm

From: Maurizio Boscarol <maurizio@usabile.it>
Date: Wed, 09 Nov 2005 18:57:24 +0100
Message-ID: <43723884.3030006@usabile.it>
To: "Bailey, Bruce" <Bruce.Bailey@ed.gov>
CC: w3c-wai-gl@w3.org

Bailey, Bruce wrote:

>Awesome job, thanks Gregg.  And Wendy, the supports documents are excellent too.
>
>I recognize that not all the reasons listed were substantive enough to be included as to why validity should be Level.  However, I believe these two should be included:
>
>22.  The evidence is that the one-way statistical correlation between sites that are valid to those that are accessible is too overwhelming to be explained away by the hypotheses offered.
>  
>


This is totally undemonstrated as far as we know.. There is no 
statistical evidence I know, if you have one, we should see the data.

But even if you have, that wouldn't be enough. There's a lot of 
methodological problem in that statement:

a. How do you measure an accessible page? According to wcag 1.0 AA or 
higher? Then, of course, the correlation is given by the tools you used 
to measure accessibility, because AA page must be valid! This has no 
meaning: it's a tautology.

b. Accessibility should be measured with and independent measurement (on 
which type of measurement scale?), as the capability of user to access 
and successfully use the page, and then observe if the page is valid or 
not. Only in that case, with a double-blind procedure, you could test 
the correlation.

c. Even with a high difference between groups (valid and invalid, with a 
measurement of empirical accessibility), you can't always assert a 
causal relation. Both validity and accessibility could be caused by 
same, external factors. I.E., having meaningful textual equivalent, 
having good structured and readable datas, etc. This factors are the 
component relevant to accessibility, of course, and that's what 
guidelines are supposed to check!

d. What do you mean by putting together "One-way" (often referred to 
anova, sometimes to correlation), correlation and hypotesis? I'm a bit 
confused on the evidence you have. Can you explain what you exactly mean?

So I suggest to go slowly, before accepting this argument as a "pro"... 
Then if we had, it would be an interesting argument, of course.


On the con- side.

I think the most evident argument not to put validity at level 1 is the 
first: "Validity is not required for accesssibility. An accessible 
website can still contain invalid code". If this is true, as we already 
agreeed, we'd committing a certain level of error by letting potential 
accessible pages out from a test based on our set of guidelines.

Precisely, we'd be committing an error of 2nd type: we're refusing a 
site as inaccessible when it's accessible. This type of error is 
strongly related to the power of the test. We should estimate how often 
we would commit that error (by an empirical test with a large set of 
real sites and a large number of disable user: impratical, I fear). If 
the probability is too high, we're making up a bad tool.

Well, that's a technical discussion.

Maurizio
Received on Wednesday, 9 November 2005 17:45:32 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:47:40 GMT