Comments on: The 96.97 percent problem http://www.robweir.com/blog/2006/08/9697-percent-problem.html?utm_source=rss&utm_medium=rss&utm_campaign=9697-percent-problem Thinking the unthinkable, pondering the imponderable, effing the ineffable and scruting the inscrutable Tue, 07 Feb 2012 11:20:47 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: Rob http://www.robweir.com/blog/2006/08/9697-percent-problem.html#comment-355 Rob Wed, 24 Jan 2007 03:17:00 +0000 http://www.robweir.com/blog/2006/08/the-96-97-percent-problem.html#comment-355 Hi LB, <br /><br />I have no doubts that my web site traffic is atypical. I never claimed it was. My point is that the number of hits you get (whether tens of thousands or millions) doesn't matter so much as the way you do your sampling. <br /><br />It is well-known to survey practitioners that a survey of only 400 people can represent the opinions of the entire country within 5%, if you pick a unbiased representative sample. Google for "survey of 400" to see how often that magic number is used.<br /><br />Similarly, a survey of 10 million people can be useless if it is not representative.<br /><br />Also, I'm not old. Hi LB,

I have no doubts that my web site traffic is atypical. I never claimed it was. My point is that the number of hits you get (whether tens of thousands or millions) doesn’t matter so much as the way you do your sampling.

It is well-known to survey practitioners that a survey of only 400 people can represent the opinions of the entire country within 5%, if you pick a unbiased representative sample. Google for “survey of 400″ to see how often that magic number is used.

Similarly, a survey of 10 million people can be useless if it is not representative.

Also, I’m not old.

]]> By: LB http://www.robweir.com/blog/2006/08/9697-percent-problem.html#comment-354 LB Wed, 24 Jan 2007 02:21:00 +0000 http://www.robweir.com/blog/2006/08/the-96-97-percent-problem.html#comment-354 Whether or not the last digit in "0.36%" is significant or not is not important.<br /><br />Of course Rob is entittled to his opinion. Sampling bias also applies to Rob's own site. We don't know if Rob's cross section of hits represents the Web as whole either. Who knows, maybe Rob is lying or maybe Rob has lots of friends who use FF.<br /><br />The real difference is that OnesStat sampled 2 million hits while old Rob had a mere 37,000 hits. Businesses pay cold hard cash for OneSat's services while old Rob runs a blog for free.<br /><br />You can believe whoever. I trust OnesStat over old Rob. Whether or not the last digit in “0.36%” is significant or not is not important.

Of course Rob is entittled to his opinion. Sampling bias also applies to Rob’s own site. We don’t know if Rob’s cross section of hits represents the Web as whole either. Who knows, maybe Rob is lying or maybe Rob has lots of friends who use FF.

The real difference is that OnesStat sampled 2 million hits while old Rob had a mere 37,000 hits. Businesses pay cold hard cash for OneSat’s services while old Rob runs a blog for free.

You can believe whoever. I trust OnesStat over old Rob.

]]> By: Anonymous http://www.robweir.com/blog/2006/08/9697-percent-problem.html#comment-85 Anonymous Sun, 17 Sep 2006 18:57:00 +0000 http://www.robweir.com/blog/2006/08/the-96-97-percent-problem.html#comment-85 Man, are some of your readers dumb or what? Scary that some people just will never be able to get it. Really scary. Oh, look out my window, there is the result; now I get it. Heh. Man, are some of your readers dumb or what? Scary that some people just will never be able to get it. Really scary. Oh, look out my window, there is the result; now I get it. Heh.

]]> By: Rob http://www.robweir.com/blog/2006/08/9697-percent-problem.html#comment-61 Rob Tue, 22 Aug 2006 16:08:00 +0000 http://www.robweir.com/blog/2006/08/the-96-97-percent-problem.html#comment-61 No, I did not write an entire blog post just about a decimal point.<br/><br/>I was illustrating a point about precision and accuracy. The numbers in the OneStat press release exhibited what is called "false precision", giving the appearance of more precision than warranted by the data. To the average reader, this false precision also carries with it an implication of high accuracy, which is not necessarily true. The accuracy of the survey is not substantiated. We were given no indication that the survey was representative of "global use", although the results claimed to demonstrate that. I also gave an example where I could legitimately calculate a high precision answer that Linux has a 21.1% market share.<br/><br/>The point is a company can have a slick web site, issue a slick press release with a bunch of numbers, have that press release copied and quoted all over the world, even have the word "stat" in their name and still come up with a survey that would get a failing grade in a intro to statistics class at a community college. I don't blame them. I blame everyone else. 98.3432% of people simply live in awe of numbers, and don't question them. No, I did not write an entire blog post just about a decimal point.

I was illustrating a point about precision and accuracy. The numbers in the OneStat press release exhibited what is called “false precision”, giving the appearance of more precision than warranted by the data. To the average reader, this false precision also carries with it an implication of high accuracy, which is not necessarily true. The accuracy of the survey is not substantiated. We were given no indication that the survey was representative of “global use”, although the results claimed to demonstrate that. I also gave an example where I could legitimately calculate a high precision answer that Linux has a 21.1% market share.

The point is a company can have a slick web site, issue a slick press release with a bunch of numbers, have that press release copied and quoted all over the world, even have the word “stat” in their name and still come up with a survey that would get a failing grade in a intro to statistics class at a community college. I don’t blame them. I blame everyone else. 98.3432% of people simply live in awe of numbers, and don’t question them.

]]> By: Anonymous http://www.robweir.com/blog/2006/08/9697-percent-problem.html#comment-60 Anonymous Tue, 22 Aug 2006 14:45:00 +0000 http://www.robweir.com/blog/2006/08/the-96-97-percent-problem.html#comment-60 Ok, so instead of 0.36%, Linux usage is actually 0.3% (+/- 0.1%). Is that your point? I´m impressed. Ok, so instead of 0.36%, Linux usage is actually 0.3% (+/- 0.1%). Is that your point? I´m impressed.

]]> By: Anonymous http://www.robweir.com/blog/2006/08/9697-percent-problem.html#comment-57 Anonymous Tue, 22 Aug 2006 05:52:00 +0000 http://www.robweir.com/blog/2006/08/the-96-97-percent-problem.html#comment-57 That's some cool shit. That’s some cool shit.

]]>