Extreme events: blending PCA and ICA [15]

Go to: Summary | Previous | Next   
Bullet points include: Suppose we think that ‘behaviour’ that is highly non-Normal is likely to be ‘interesting’ (i.e. worth exploring further) and probably ‘meaningful’. Suppose we also associate non-Normality with (excess) kurtosis. Conveniently: All linear combinations of independent distributions have a kurtosis less than or equal to the largest kurtosis of any of the individual distributions. Kurtosis is scale independent (i.e. k.x has the same kurtosis as x if k is a scalar)

Contents | Prev | Next | Library

Desktop view | Switch to Mobile