g******e 发帖数: 352 | 1 书上一般都说leptokurtic是more peaked, fatter tails
但网上的这张图上却明明是leptokurtic 是thin tail
我总感觉platykurtic 因为less peaked, pdf更宽,也会使得离mean远的点的prob变大
而导致fat tail.
大牛说说到底该怎么理解?有包子答谢 |
p*****k 发帖数: 318 | 2 (plz leave baozi to others)
it's usually hard to see on a linear scale - try the log-plot here:
http://en.wikipedia.org/wiki/Kurtosis#Kurtosis_of_well-known_distributions
your intuition is correct for Gaussian case: wider->fatter tail.
however, there is an additional intersection for a non-zero-kurtosis
p.d.f. with Gaussian p.d.f. |
g******e 发帖数: 352 | 3 谢谢,wiki的图还是精确多了
看起来platykurtic的pdf在sigma 到 2 sigma这段比gussian fat,
但再远到真正tail的地方就又比guassian低了.
【在 p*****k 的大作中提到】![](/moin_static193/solenoid/img/up.png) : (plz leave baozi to others) : it's usually hard to see on a linear scale - try the log-plot here: : http://en.wikipedia.org/wiki/Kurtosis#Kurtosis_of_well-known_distributions : your intuition is correct for Gaussian case: wider->fatter tail. : however, there is an additional intersection for a non-zero-kurtosis : p.d.f. with Gaussian p.d.f.
|
s***e 发帖数: 267 | 4 Leptokurtic is more peaked does not conflict with fat tails. Basically you
have larger area around the peak as
well as far down the tail, but have much smaller area in between.
The graph you draw is a little bit misleading as it does not give a clear
picture about the tail behavior. If you
plot a normal together with a laplace distribution you will see the
differences.
【在 g******e 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 书上一般都说leptokurtic是more peaked, fatter tails : 但网上的这张图上却明明是leptokurtic 是thin tail : 我总感觉platykurtic 因为less peaked, pdf更宽,也会使得离mean远的点的prob变大 : 而导致fat tail. : 大牛说说到底该怎么理解?有包子答谢
|
t**o 发帖数: 64 | 5 这网上的图的 variance 没有 normalize,画错了
【在 g******e 的大作中提到】![](/moin_static193/solenoid/img/up.png) : 书上一般都说leptokurtic是more peaked, fatter tails : 但网上的这张图上却明明是leptokurtic 是thin tail : 我总感觉platykurtic 因为less peaked, pdf更宽,也会使得离mean远的点的prob变大 : 而导致fat tail. : 大牛说说到底该怎么理解?有包子答谢
|