Comments on: Rankings: never a bell curve, not always a power law http://www.econometa.com/archives/25 The economy of stuff about stuff Fri, 21 Nov 2008 00:48:54 +0000 http://wordpress.org/?v=2.2.1 By: admin http://www.econometa.com/archives/25#comment-15776 admin Thu, 19 Jul 2007 23:02:52 +0000 http://www.econometa.com/archives/25#comment-15776 Hi Michael, Thanks for the comment! You're right, how sharply the tail ends and how peaked it is at the head can completely change the meaning of a ranked histogram. That's why they can be misleading; it's safer to consider the associated PDF instead, if the question you want to answer has to do with averages, etc. Hi Michael,

Thanks for the comment! You’re right, how sharply the tail ends and how peaked it is at the head can completely change the meaning of a ranked histogram. That’s why they can be misleading; it’s safer to consider the associated PDF instead, if the question you want to answer has to do with averages, etc.

]]> By: michael webster http://www.econometa.com/archives/25#comment-15774 michael webster Tue, 17 Jul 2007 18:47:58 +0000 http://www.econometa.com/archives/25#comment-15774 I found this discussion very helpful, especially in conjunction with reading Taleb's new book. The histogram that you derived from a gaussian PDF actually does look like many of the reall histograms -with a sharp ending marking the end of the shelf, lack of any new entries, etc. Interesting that the difference between a gaussian PDF, the related histogram, and histograms with power law PDF's may just depend upon how sharply the tail comes to an conclusion -or at least that is my conclusion after reading this article. I found this discussion very helpful, especially in conjunction with reading Taleb’s new book.

The histogram that you derived from a gaussian PDF actually does look like many of the reall histograms -with a sharp ending marking the end of the shelf, lack of any new entries, etc.

Interesting that the difference between a gaussian PDF, the related histogram, and histograms with power law PDF’s may just depend upon how sharply the tail comes to an conclusion -or at least that is my conclusion after reading this article.

]]> By: purple motes » lack of power laws and other popularity problems http://www.econometa.com/archives/25#comment-4356 purple motes » lack of power laws and other popularity problems Mon, 21 Aug 2006 23:25:42 +0000 http://www.econometa.com/archives/25#comment-4356 [...] Power laws and log-normal distributions are two-parameter distributions. The distributional form that best characterizes traffic to all pages may have many more than two parameters. Compared to a log-normal distributions and other distributions with more than two paramters, an approximating line has greater value for providing a simple, intuitive description of an important part of the popularity distribution. [...] […] Power laws and log-normal distributions are two-parameter distributions. The distributional form that best characterizes traffic to all pages may have many more than two parameters. Compared to a log-normal distributions and other distributions with more than two paramters, an approximating line has greater value for providing a simple, intuitive description of an important part of the popularity distribution. […]

]]> By: Adam http://www.econometa.com/archives/25#comment-114 Adam Mon, 31 Oct 2005 18:47:21 +0000 http://www.econometa.com/archives/25#comment-114 Hi Student, I'm not sure what you mean about the "general shape" of a CDF. Maybe you mean that a CDF often gives the probability of any outcome less than or equal to x, while I'm showing the probability of any outcome *greater* than or equal to x? If so, this is not an unusual variant, as I mentioned at the very end of <a href="http://www.econometa.com/archives/15" rel="nofollow">this post</a>; for example, a Pareto distribution is this latter type of CDF. Adam Hi Student,

I’m not sure what you mean about the “general shape” of a CDF. Maybe you mean that a CDF often gives the probability of any outcome less than or equal to x, while I’m showing the probability of any outcome *greater* than or equal to x? If so, this is not an unusual variant, as I mentioned at the very end of this post; for example, a Pareto distribution is this latter type of CDF.

Adam

]]> By: Student http://www.econometa.com/archives/25#comment-111 Student Sun, 30 Oct 2005 19:42:28 +0000 http://www.econometa.com/archives/25#comment-111 Dear poster, Please revise on the general shape of a CDF before posting graphs. It will make your point palatable beyond the statistically illiterate. Cheers, Student Dear poster,

Please revise on the general shape of a CDF before posting graphs. It will make your point palatable beyond the statistically illiterate.

Cheers,

Student

]]>