Hi Graziella!

Sergei makes an interesting example. A more general example, in my opinion, would relate to a panel set-up (e.g. of firms). Think for instance of macro-shocks affecting the noise component of a production function. I expect the skewness of this noise component could vary through time following macro-cycilcal patterns,

Hope this helps

S.

Webpage: http://www.unisa.it/docenti/sergiopietrodestefanis/index;

Repec Papers: http://econpapers.repec.org/RAS/pde148.htm;

SSRN Papers: http://ssrn.com/author=451445

from the point of view of data analysis, which could be the existence of latent variables, or inadequate data stratification,

it is difficult to give in my case a better assessment in the context of the analysis is unknown.

Let an experimenter want to record the failure times of some experimental units. In this phenomena one can record the time of failure exactly at the failure or after a little moment, so in this case -ive error never occur only +ive error possible. Therefore, error in this case can't be symmetric.