Re: std dev v. std err of est?

>>Second, is there a commonly used abbreviation for the
>>std error of the est? For example, sd is commonly
>>use for the std dev. I find SEE in a google search.
>>Is that in common usage?
>
>The common terms for s in a regression is std err of estimate (SEE) or
>std error of the regression (SER). Both are used.

Unfortunately, this is true. Despite their being illogical terms.

> s is an estimate of
>the standard deviation of the error terms in the regression. The
>informal convention is to call an estimated standard deviation a
>standard error.

Only in the illogical world of traditional regression terminology.

Elsewhere in statistics, a "standard error" is the estimated standard
deviation of the estimator for some parameter, which is an indication
of how accurate the estimate is. We expect standard errors to get
smaller when we get more data. The standard deviation of the
residuals in a regression model is ITSELF a parameter, not an
estimator. One would not expect this so-called "standard error of the
estimate" to get smaller if you got more data, proving that it's not a
"standard error" at all.

The logical terminology is to just call it the "estimated residual
standard deviation".

At least you're focusing on the right thing, though. The standard
deviation of the residuals is much more meaningfull than R^2. But
if you do use R^2, please don't refer to it by that preposterous
term, "coefficient of determination"...

----------------------------------------------------------------------------
Radford M. Neal radford@xxxxxxxxxxxxxx
Dept. of Statistics and Dept. of Computer Science radford@xxxxxxxxxxxxxxxxxx
University of Toronto http://www.cs.utoronto.ca/~radford
----------------------------------------------------------------------------
.