[Informal Meeting!]

WCAG 2.0 Evaluation Methodology Task Force Teleconference

01 Aug 2013

See also: IRC log


Liz, Shadi, Kathy, Peter, Vivienne, Peter
meeting was canceled on short notice


<MartijnHoutepen> Just received e-mail from Eric, he's cancelled this meeting

<MartijnHoutepen> His phone and internet are both down

Liz: like the proposed text for confidence
... don't think can have more rigor

Shadi: no meeting next two weeks
... but watching the list for emails
... putting together an updated Editor Draft
... please engage in online discussions

Vivienne: have student working on random sampling generator

Liz: Peter suggested pattern recognition

Vivienne: forwarded suggestions to PhD supervisor

Liz: for more complex applications it may need more "intelligence"

Peter: did some research on statistical sampling methods
... looking at statistical rigor and certainty
... relating to sample size and variance
... several examples looked at polling groups of data
... including statistical significance and margin of error
... quite a lot of knowledge that could be productively applied
... that would be a good thing to bring in if a student has the skills
... tool could support "knowledge" about the website
... concerned about how this applies to web applications
... have an example of an application that has a save button that appears on several pages
... pressing that button inserts a link with an accessibility error
... not able to predict this behavior
... can only talk about what was identified and tested

Vivienne: another example I have is where an application is actually a complete process
... so actually can't sample but need to test all steps

Peter: agree that not accessible if step cannot be accomplished
... but what if one error appears somewhere on these steps
... does that prevent a person from using the application?
... WCAG does not distinguish the severity of violations
... but in pratice it may sometimes not be a problem
... need to make the report useful in the real world
... to deliver this type of information

Vivienne: agree, sometimes errors have varying levels of impact
... Roger Hudson worked on a "calculator" to help prioritize and identify the major errors

<Vivienne> Roger Hudson's tool can be downloaded from: http://usability.com.au/2013/01/accessibility-priority-tool/

Shadi: model of "minimum process" to assess a website with sufficient confidence
... then additional requests by the Commissioner

Peter: probably need to distinguish between static websites and web applications
... no clear line between these

Vivienne: agree with Peter
... more likely to miss specific interactions in web applications

Shadi: back to the random sampling generator
... if the tool has more "knowledge" about the website, isn't that then structured sampling?

Peter: the more you know about the website, the smaller sample size you need to maintain the same margin of error

<korn> See http://en.wikipedia.org/wiki/Sampling_(statistics)#Sampling_methods

<korn> See "Systematic sampling", which appears to be the term

<korn> Also "Stratified sampling"

Shadi: random sampling was intended as a back-check the structured sampling

Peter: question in stratified sampling is to know what you may have missed
... interesting article above

<korn> See "Practical example" at http://en.wikipedia.org/wiki/Stratified_sampling

Vivienne: tool is designed to generate a purely random sample
... without prior knoweldge
... think Peter is suggesting enhancements
... but may not be possible within this student project

Liz: what Peter suggests is not that hard
... if one can come up with classifications for the pages
... for example pages with tables, forms, etc

Peter: classifying pages then sampling from those is probably achievable
... but might not be what we want
... for example for pages that have tables, many may have been produced using template A and others using template B
... need to ensure not to miss these few pages

Vivienne: fabulous feedback

Liz: instead of stratified, can call it "choice based"
... choose the classes then it will sample from them

Shadi: remember a project looking at page elements and calculating "indicators" from that
... from those they created "clusters" of "similar" web pages
... and then selected from each of these clusters representatives

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.138 (CVS log)
$Date: 2013/08/25 08:36:57 $