Re: [Gnuplot-bugs] fit.c duplicate extimation of error?

[Hans-Bernhard B. <br...@ph...>]

On Mon, 5 Apr 2004, Isao Sakane wrote:

> In fit.c: The covar matrix is calculated base on err_data if given by a 
> user. Thus, scaling of parameter errors occurs twice and cancels out 
> each other if err_data are explicitly given.

Exactly.  And that's done on purpose.

> "Asymptotic Standard Error" of parameters should increase as errors 
> increase.

Only if you believe the error bars more than the deviations of the
data from the model.  In cases where the WSSR/ndf is far away from
its expected value of 1.0, the physical interpretation is that
either the errors are nonsensical, or the model isn't usable.

Scaling around the parameter errors doesn't make very much sense in
such a situation.

-- 
Hans-Bernhard Broeker (br...@ph...)
Even if all the snow were burnt, ashes would remain.