Should we add functions like mae, sse, mse, rmse here?

[singularitti]

Definition: https://www.sfu.ca/sasdoc/sashtml/ets/chap30/sect19.htm

I know these functions are very easy to be implemented with mean, abs2, sqrt, etc. But it would be really nice to have them predefined in Julia.

[ajinkya-k]

Some of these seem to be implemented in StatsBase.jl, see here

[nalimilan]

Yes if we add these it will be in StatsBase for consistency with existing functions. I have transferred the issue. But note that some of the functions you mention actually already exist with another name. Do you still think some are missing?

[ajinkya-k]

@nalimilan I think one useful thing to add would be the degree of freedom corrected versions. For instance the msd and rmsd use the length of the input vectors as the denominator though sometimes we might want to divide by n - 1 instead. Maybe we can call that the mse / rmse

[ajinkya-k]

Also I think it might be useful to add aliases sse for the sum of squared error mae for mad

[nalimilan]

I guess we could add a corrected argument to these functions, similar to what we have for cor. Calling these mse and rmse would be confusing as there's no special reason why these should be corrected and not msd and rmsd AFAICT.

Concerning aliases, I'm hesitant to add these as it makes the API more complex and I don't think we do that usually. Feel free to make a PR to mention the alternative names in the docstrings though, that should make it easier to find the functions.

[ajinkya-k]

You're right. My suggestions were mostly just aliases. From my experience people in statistics mostly use rmse instead of rmsd though it might be specific to the people/school I work with