Moving forward with standard deviation and standard error function names.

[Nosferican]

Currently StatsBase has std for standard deviation and used stderr for standard error. Given the decision at Base to use stderr as one of the three input/output (I/O) connections, StatsBase must decide how to rename it.

Possible candidates include:

• sd and se
• stddev and stderror

Relevant discussions occurred in julia PR#26636, Slack #statistics, and #368.

[ararslan]

+1 for stddev and stderror. I think we should favor more descriptive names wherever possible. Also note that there's likely a lot of code that currently uses sd and se as variable names, which would then be broken by claiming those names for the functions themselves.

[Nosferican]

Those are two very valid arguments.

stderror(model)

would still be a bit shorter than

sqrt.(diag(vcov(model)))

which is at least the longest the standard error function should be.

[nalimilan]

Why not, but if we rename std to stddev, we should do the same in Base.

[stevengj]

My suggestion would be to leave std as-is, and use stderror for the standard error.

The standard deviation is more common, std is a well known name for it that hasn't caused any problems so far, and std mirrors var. Also, leaving std (and var) as-is would be less disruptive.

[ararslan]

Sure, that makes sense to me.

[nilshg]

This is probably decided now so can be closed?