"If you torture the data long enough, it will confess to anything" nghĩa là gì?


"If you torture the data long enough, it will confess to anything" = tra tấn dữ liệu đủ lâu thì tự khắc nó sẽ thú nhận -> nghĩa là số liệu, thống kê đều có thể bị thao túng để phục vụ mọi tình huống, mục đích.

Ví dụ
20th-century British economist and Nobel laureate (người trúng giải thưởng) Ronald H. Coase posited, “If you torture the data long enough, it will confess to anything.” The dilemma (tình trạng khó xử) we face as analysts (a generous title, in my case) is how to filter (lọc) through the raw data to provide real insight. There’s a difference between eliminating noise and pushing a narrative.

As noble laureate Ronald Coase once said, “If you torture the data long enough, it will confess to anything.” Overfitting is malpractice (phi pháp) in which models are trained to fit exactly against the existing data but cannot perform accurately against unseen data.

Nobel Prize winning economist Ronald Coase is credited to have said: “If you torture the data long enough, it will confess to anything”. Indeed, evidence suggests that all disciplines (kỉ luật) of scientific literature are not free from bias (thiên vị) and questionable practices such as selectively reporting hypotheses, excluding data points and variables post-hoc, and rounding “p value” - so non-significant results (arising by chance) become significant (that is, has a systematic effect).

Ngọc Lân

Post a Comment

Tin liên quan

    Tài chính

    Trung Quốc