sassa_nf ([personal profile] sassa_nf) wrote in [personal profile] juan_gandhi 2020-02-05 06:52 pm (UTC)

Thanks for the writeup.

"они берут пул моделей (27 штук, емнип) и на каждом годе делают вид, что у них 27 случайных величин, находят среднее, стандартное отклонение и все это рисуют на графиках... Эти величины и играют роль "ошибки""

That's an odd way to calculate errors.


"Т.е. мало того, что выдачи моделей не имеют отношения к iid (independent and identically distributed) случайным величинам"

Ok. That's an interesting methodology.

Rainfall modelling used literal deviations of observations from model predictions as errors. That is, e_i = o_i - f(i) - observed value take away the function output. (And yes, the function was full of things like literally the angle of incidence of sun rays, Boltzmann law for evaporation, etc)

Post a comment in response:

This account has disabled anonymous posting.
If you don't have an account you can create one now.
HTML doesn't work in the subject.
More info about formatting