Amazon Sues Alexaholic

theodp writes "ZDNet reports that as Jeff Bezos tap-danced out of a cringe moment at Web 2.0 Expo prompted by Tim O'Reilly's questioning of why Amazon couldn't get along with Alexaholic (now Statsaholic), Amazon had already filed a lawsuit to legally spank the tiny company into oblivion."

Comments

Reasons to like Alexa?

[drgonzo59]

From an excerpt in the article titled "Reasons to like Alexa" a response to the claim that Alexa's data is not representative was Statistical significance is attainable with only a small subset of the population - ask a pollster or a high school math teacher..

That is a mistake, or rather a mistaken response to the claim. Yes, statistical significance is attainable but only if the sample is representative (i.e.) is random. The critics' claim is that Alexa's data is not representative, in other words the sites that choose to give Alexa their data are somehow don't represent a random sample of all the websites out there. It isn't a question of size but rather of quality.

Re:Reasons to like Alexa?

[Assassin bug]

True, representation is reliant on how the samples are obtained and the response variable used for the estimate. However, representation and the randomness of the data are not necessarily related. There are different "kinds" of random-sampling techniques (e.g., systematic or arbitrary). Also, the data itself has its own measure of randomness. You can have a non-random, representative data set. You can even have non-random data with heterogenious variance and have it be representative. What matters, in statistics, is that the assumptions for whatever statistical test is used are checked and that corrections to the analysis are made to accomodate for violations of the assumptions.

Re:Reasons to like Alexa?

[suv4x4]

Yes, statistical significance is attainable but only if the sample is representative (i.e.) is random.

Actually "random" would be the opposite of "representative", as long as statistics are concerned. Represenative means the same proportions of the subgroups in the samples are the same as the whole. The subgroups should be carefully chosen to represent properly what could bias or change the outcome of the results.

As an extremely simple example, you want in the sample to have the same proportions of age, gender, income, professions etc (some of those categories may not matter in certain studies).

Re:Thanks Tim

[Shemmie]

http://www.tomrafteryit.net/oreilly-trademarks-web -20-and-sets-lawyers-on-itcork/

I seem to remember hearing this, way back when.

Re:Thanks Tim

[maxume]

http://www.tomrafteryit.net/sorry-tim/

In short, O'Reilly is partnered with CMP and CMP has the mark and sent the letter.

He was screen scraping...

[xENoLocO]

...not using the APIs.

He was "avoiding an API fee", but the data he wanted was not available through the API anyways, so he screen scraped alexa. If alexa had wanted that data available they would have made it available through the API.

The guy (hornbaker) admittedly says he wants to turn this into a PR battle. And I remember him explicitly trying to stick it to amazon before he changed the site name.

I don't really know who the hell to cheer for here, so I'm just gonna sit back and watch.