Re: [Fis] 2004 FIS session: concluding comments

Replying to Guy -- Ah, well, I am not too surprised at hearing this. At
least two authors in the past (including Mandelbrot) have argued that power
laws can be viewed as transformations of LOGnormal distributions (which
also have long tails). However, on one list, after I stated this, someone
(I can't find the posting) argued that this mapping is superficial. That
is, that data that can be plotted as a power law are data that, looked at
another way, would be shown to be lognormally distributed.

STAN

>Dear Stan et al.,
>
>On Jun 26, 2004, at 2:32 PM, Stanley N. Salthe wrote:
>
>> Pedro said:
>>
>>> Perhaps it is in this very context where we should consider the need
>>> of
>>> more fine-tuned conceptualizations on information /entropy. For
>>> instance my
>>> mild criticisms on Shannon's overextensions in last posting (Aleks was
>>> right saving the cleanliness of his formalism assumptions) and
>>> 'power law'
>>> inspired approaches to entropy such as Tsallis. In network analysis,
>>> something similar has already happened ---Barabasi and Albert,
>>> working on
>>> real-world networking. Scale-free nets' signatures can be found in
>>> the market and companies distributions (Internet structure too), in
>>> nervous systems, in protein networks... curiously, the very info
>>> gateways I
>>> was mentioning above (and the old fis theme of info 'model systems':
>>> cells,
>>> brains, companies). It is also curious that previous to Barabasi and
>>> Albert
>>> papers, research on networking was classically dominated by Paul
>>> Erdos and
>>> Alfred Reny's views on random networks (mandating a Poisson
>>> distribution in
>>> the links). The paper I mentioned last week by Dante Chiavo was
>>> advancing
>>> not so disparaging views (with his 'multiscale entropy').
>> SS: Well, I have the sense that these scale-free, power law, 1/f
>> distributions, since they seem to be found partout (in data from both
>> material and immaterial realms) are not specfically informative of
>> anything. Likely they are artefacts of analysis!
>
>I heard John Doyle make an interesting argument about power laws at the
>ICCS 2004 conference this summer. He argued that power law
>distributions should be the ones called "normal" and that they are
>actually the distributional pattern predicted by the Central Limit
>Theorem, rather than the bell shaped curves we traditionally call
>"normal." If he is correct (the argument sounds right to me), then
>power laws are not merely artifacts of analytical methods. Instead,
>they would be ubiquitously generated by nature.
>
>Thoughts?
>
>Guy Hoelzer
>
>Department of Biology
>University of Nevada Reno
>Reno, NV 89557
>
>Phone: 775-784-4860
>Fax: 775-784-1302
>
>_______________________________________________
>fis mailing list
>fis@listas.unizar.es
>http://webmail.unizar.es/mailman/listinfo/fis

_______________________________________________
fis mailing list
fis@listas.unizar.es
http://webmail.unizar.es/mailman/listinfo/fis
Received on Mon Jun 28 22:32:41 2004