(Meta-)Erros Tipo I e II

De Netter, Marks R. An Applied Statistician’s Creed (publicado em Journal of the Royal Statistical Society, Series C, v.45, n.4, vem a tabela mais engraçada da Estatística (até agora, para mim).

errotipoiii

Agora imagine os erros tipo I e II sobrepostos nesta divertida tabela. ^_^

É a mudança climática um passeio aleatório?

Eli Dourado:

The human brain is “hard-wired” for pattern recognition. So much so, in fact, that it often finds patterns that are not really there. This is why the field of statistics is so counter-intuitive for so many people (and you thought it was all the Greek letters).

I was thinking and reading about global warming this morning when a thought struck me: could I test climate trends against a null hypothesis of a random walk? It took only a few minutes to discover that a pair of authors from the US Geological Survey have already done so. In a 2005 article in Geophysical Research Letters, Timothy Cohn and Harry Lins were unable to reject the hypothesis that the warming of the last 100 years was due to randomness. The trend is statistically insignificant.

I draw two conclusions. First, politicians should not do anything drastic to “fix” global warming. Second, if we insist on democratic governance, we ought to emphasize statistics in the high school curriculum. I propose dropping trigonometry (which is useless to almost everyone) and replacing it with statistics.