>He specifies that SD is reasonable as a mathematical tool, but not for inference about society, finance, etc.
Yes, and as SD is widely used in those so this is quite a bold statement to make an Taleb doesn't give one example of things going wrong with it and his proposal working better.
This observation is also wrong and I pointed out some areas where standard deviation is useful.
>Your comment appears to have an agenda or ideology underlying it.
Yes it's ideology of calling people out when they assert incorrect things in condescending tone without any arguments while insulting whole groups of scientists and statisticians.
>That is pretty much what his entire article is about: taking squares may not be the best idea universally.
No, article isn't about it. It was in my comment. Taleb doesn't mention how MAD works better than SD in some situations and why or why it should be substituted. Taleb doesn't point one situation when MAD works better than SD (I pointed out some where SD works better than MAD) he also doesn't address obvious problems which such substitutions create.
Heck, he doesn't even discuss how nature of SD is different than MAD, he just says they are different and the latter is more applicable for "real life".
Also he doesn't argue that MAD is sometimes more useful than SD. He argues that the latter should be disposed of outside some very narrow theoretical applications.
> Taleb doesn't point one situation when MAD works better than SD (I pointed out some where SD works better than MAD) he also doesn't address obvious problems which such substitutions create.
Ummm, are we both discussing the same article?
> Yes it's ideology of calling people out when they assert incorrect things in condescending tone without any arguments while insulting whole groups of scientists and statisticians.
No. It is the ideology of dissing Taleb. Obviously I am not his paid spokesperson. But please argue to the merit of his arguments.
To quote from Taleb's edge.org article:
>> 1) MAD is more accurate in sample measurements, and less volatile than STD since it is a natural weight whereas standard deviation uses the observation itself as its own weight, imparting large weights to large observations, thus overweighing tail events.
Also he alludes to his paper with Goldstein. It is clear form Goldstein and Taleb's manuscript that that Taleb is not just throwing these arguments to talk trash about practitioners of statistics. They report findings of an experiment with multiple groups of applied statisticians. I'll quote from it [1]:
"We first posed this question to 97 portfolio managers,
assistant portfolio managers, and analysts employed by investment management companies who were taking part in a professional seminar. The second group of participants comprised 13 Ivy League graduate students preparing for a career in financial engineering. The third group consisted of 16 investment professionals working for a major bank. The question was presented in writing and explained verbally to make sure definitions were clear."
That makes Taleb's edge.org claims far from unsubstantiated.
[1] Goldstein, Daniel G. and Taleb, Nassim Nicholas, We Don't Quite Know What We are Talking About When We Talk About Volatility (March 28, 2007). Journal of Portfolio Management, Vol. 33, No. 4, 2007. Available at SSRN: http://ssrn.com/abstract=970480
Yes.
He doesn't mention the fact that MAD loses information about volatility. He ignores the fact because that would make his whole article look silly.
>But please argue to the merit of his arguments.
It's kinda difficult when he is not saying anything specific. When he says: "look, those scientists don't grok basic math but fail for catchy names instead" without any facts, examples etc. all I can do is call him out on this nonsense. When he argues for abandoning SD I can give situations where it's not gonna work and I did.
>1) MAD is more accurate in sample measurements, and less volatile than STD since it is a natural weight whereas standard deviation uses the observation itself as its own weight, imparting large weights to large observations, thus overweighing tail events.
But this is nonsense. It's like saying measuring temperature is better than measuring mass. Those are just different things to measure and saying one is less volatile isn't really meaningful. SD contains information about volatility, MAD doesn't. That the reason SD is used for many things. When you want to substitute one with the other you gotta address how you handle that lost information.
>Also he alludes to his paper with Goldstein. It is clear form Goldstein and Taleb's manuscript that that Taleb is not just throwing these arguments to talk trash about practitioners of statistics. They
This paper wouldn't pass peer review. What was the methodology ? How much time they have ? Did they have access to a computer ?
Even if it was really serious experiment it's just toy example of people getting question asked in tricky way wrong. What about actual mistakes they make in real world because of using SD instead of MAD ? This is what he claims is a problem.
He claims SD should be abandoned in favor of MAD, what is one situation which people would get better if they do it ?
I am mocking him because is a master of using many words without making a point and somehow is good at seducing readers to like him.
Yes, and as SD is widely used in those so this is quite a bold statement to make an Taleb doesn't give one example of things going wrong with it and his proposal working better. This observation is also wrong and I pointed out some areas where standard deviation is useful.
>Your comment appears to have an agenda or ideology underlying it.
Yes it's ideology of calling people out when they assert incorrect things in condescending tone without any arguments while insulting whole groups of scientists and statisticians.
>That is pretty much what his entire article is about: taking squares may not be the best idea universally.
No, article isn't about it. It was in my comment. Taleb doesn't mention how MAD works better than SD in some situations and why or why it should be substituted. Taleb doesn't point one situation when MAD works better than SD (I pointed out some where SD works better than MAD) he also doesn't address obvious problems which such substitutions create. Heck, he doesn't even discuss how nature of SD is different than MAD, he just says they are different and the latter is more applicable for "real life". Also he doesn't argue that MAD is sometimes more useful than SD. He argues that the latter should be disposed of outside some very narrow theoretical applications.