« Vintage Photos: 1938 | Main | Vintage Photos: 1939 »

Sunday, March 14, 2010


Although it's hard to state this in a simple way (and in English!), I'll try: the error in itself is intrinsic of all models and is due to the inferential process. The presence of an error does not invalidate a model as long as it possesses some specific characteristics. When this does not happen (for example, when the distribution of the error is not shaped in a given way), it means that something is wrong with the model and an improvement must be sought. BUT the flaw of most models resides in applying the wrong technique to a given problem, which happens quite frequently to those researchers whose statistical knowledge is limited.
In conclusion, the humility doesn't lie in knowing how to deal with errors but rather in being aware that modeling need a much deeper knowledge of statistics than one might think.

The comments to this entry are closed.