Welcome to Black Adder's Site
Statistically Challenged - Why the ISP Filtering Report is a FAIL
With the public release of the ISP Content Filtering Report managed by Enix Testlab, I was one of many who have downloaded and read the report.
I am not going to give a spiel about the pros and cons of the filter, but as a researcher I must take offence to the 'statistics' within the report, which are by any standards appalling.
It is in the best interests of the public to know how bad the information that the Government is basing this decision from is.
The list of statistical violations includes misrepresentative statements, misguided conclusions, and above all excluded information make the report not worth the time it takes to read.
A few highlights of the report are as follows:
On page 6:
Circumvention - “Filtering of additional categories of content enabled ISPs to implement measures which made some common circumvention techniques difficult. For example, a third party website which hides the origin of the requested content (proxy site) can be included in a wider list of URLs to be blocked.”
On page 31:
“70 percent of respondents indicated there was no feature of the service that was
The actual question from the survey (listed on page 92)
What were the features of the filtered service that you LEAST liked:
• Hard to use
• Not effective in blocking access to content I have deemed inappropriate
• Did not provide enough choice
What does that mean? Did respondents find all of the features so terrible there was no clear worst feature? What is a feature of the service, according to the question “Hard to use” is a feature.
On Page 93:
“12. Were there any unexpected consequences of using this filtered service?”
I for one would like to know what the answer to this question was.
On page 6:
“Overall the service offered by the ISPs was considered effective by customers, with
around two-thirds of customers participating in the survey indicating that they would
either probably or definitely continue using additional content filtering services.”
My big gripe with this statement, if you look at the survey on page 93, there was no probably option for participants to choose, there was only possibly or definitely, this is in effect a lie, as not one participant said they would probably consider it.
From page 6 we can see that around 2/3 of participants (I wonder why we have not got the real percentage here??) would possibly or definitely continue using the filter. This could mean that 1% would definitely continue, and 65% would possibly. To make this statistic valid, all of the numbers would need to be released.
Without the standard deviation, means, variances, and skew / kurtosis there is already little data to support the claims of this report. Now we must look at the above two statements with a critical eye.
On page 40:
“Thirty seven percent of respondents had children 15 years or younger using an
internet service that was being filtered.”
From page 40 we can see that 37% of participants had children. Assuming (possibly the only good assumption on this page, and necessitated by the lack of data in the report) that they will want their internet filtered, we can look at how many participants without children wanted the filter. If we mix the 37% with kids up with around 66% of participants possibly or definitely considering continuing filtering, at best we can see 29% of participants (46% of participant without kids).
Continuing on page 93:
Question #2 Gender. How many participants came from each sex? Was there any differences between the two sub-groups? The answers to these questions are typically able to identify any gender bias in the survey, and help support the research.
That’s all for now, but have a poke around yourself, see what you can find, there is a lot more gold to practice your statistical detective skills on! If you find some more, write it up and drop me a line. I’ll throw it up for you black69<at>tpg<dot>com<dot>au.
I'm a Kiva - Don't donate your cash, invest it in a life!
Transparent loans to those in need, I like to know where my $1 goes. Have a look!
What are his evil doings? Perhaps they are best described by Censordyne
After Racing in the Wally Wombat September 20th, Spargo Creek, this is how a good Datsun 180B SSS Sounds with only a Resonator to keep her quiet.
Datto Sound Clip 1 On the Flat.mp3
Datto Sound Clip 2 - Pulling Uphill.mp3
To the left is Gary's Online Image (Updated every 30 Seconds). To the right is a Capture just so you know which fish is Gary!
If you'd like to see Gary's Webcam Click on the Image Above, or this link.
Gary is the largest Clown Loach in my tank. He's been there for about 3 years so far.
Remember Gary has to sleep so between 16:00 and 21:00 GMT+10 is your best chance to see him.
PHP has dawned!!! Black Adders Theory of Brain GA (Genetic-algorithms)
the human brain is capable of performing a Genetic-algorithm approach to problem
solving. This is realised by your intuition. You know something and have no idea
how or why? How does a GA work? Give it a goal, more specifically a way to evaluate
itslef (good/bad) and a way to merge/splice/mutate with copies of itself. It then
goes off and evaluates->mutates->evaluates->mutates untill we get the answer.
With no logic or knowledge applied. Queer eh?
This page has been *touched* by Doopz
Black Adders site for the Targa Rally has begun