Why MNC only for normal noise?

Question

Mathematics

Why is the use of least squares for approximating time series justified only for those cases where the noise has a normal distribution?

Answer 1

The method will work better for the normal distribution of the deviation. it's built into it.
It is easy to show this - you can take some artificial distribution, for example, the deviation is always by a given constant c. Then the function will simply get the offset f`(x) = f(x) + c and approximated by LSM will remain so.
But if the distribution is center-weighted or uniform, then the method will work

Answer 2

Bullshit.
update
Mercury13 is almost right, the only requirement is no odd moments, i.e. even distribution density.
Vasily Melnikov , accordingly, drove the blizzard away, not being able to distinguish a random error (i.e. noise) from a systematic one. It happens.
d0lph1n , having answers like this to his credit , turned out to be an off-topic charlatan . It happens, but it wouldn't be better.

Answer 3

Depends on what meaning we put in the word "justified."