HELP - RMSE equivalent to sqrt[(1-R^2)] * SD(y)?

Q&A DQ Courses
#1

Hi everyone

Can someone mathematically tell me why RMSE is equivalent to this formular (Or have i misunderstood something?):

sqrt[(1-R^2)]*[SD(y)]

R^2 is the coefficient of determination and SD(y) is the standart deviation of y.

As stated in this article:

#2

The shortcut rmse formula only applies to linear regression rmse.

#4

I tried to wright detailed answer but the formulas didn’t show as it suppose to.
as R^2 is calculated as in this article: https://en.wikipedia.org/wiki/Coefficient_of_determination
so if you multiply inside the square root in rmse’s formula by sum(y_i - mean(y))^2 / sum(y_i - mean(y))^2
you should be able to see how they are equal.

#5

Hi Abdelrahman

Thanks I think I figured it out. I also replaced sdev(y) with the formular for standard deviation.
I was able to use calculus to get the formula for RMSE.

Kasper

2 Likes
#6

Ok thanks

Weird it doesn’t say in the article. The pictures show it though.

Kasper

1 Like