gnuplot-beta

[<pl...@pi...>]

Hi,

is there some way to adjust the degrees of freedom used by fit ?

This is often not the same thing as the number of data points - num. params.

For example if the data is low-pass filtered the data points are no 
longer independent and the d.f. is much less. Not accounting for this 
will give incorrect stats.


Also I'm a little confused by this:

rms of residuals      (FIT_STDFIT) = sqrt(WSSR/ndf)    : ......


The root mean square is not a statistic that depends upon d.f. , it's 
the root of mean of the squares.

Is this incorrectly labelled under the same assumption that the ndf=num 
of data pts.?

Since ndf depends upon the number of fitted params this can not be 
strictly correct anyway.

Peter.