Sorry, just to clarify my point:
The error bars on the 1st data point and on the fitted dilution factor are
not immediately related. Rather, by showing the dilution +- its error bar
(from the fit) we are able to judge the quality of asymmetry measured
for each data point. If the asymmetry measurament had problems (even if
the AF was high) the chi2 of the fit and the uncertainty on the dilution
factor will be large ALSO for large statistics.
There is another way to look at it: for the same resulting AF, the first
point of the asymmetry could be above/below the curve by the same amount,
which would confuse the comparison with presumed AF.
Anyways, look into the original show_ep_asym.C. Here is how we fit asymmetry
histogram asl[0] with some user parametrization of the theory asymmetry
// Define the user function. User function depends only on 1 scale parameter, [0]
funl = new TF1("funl","[0]*(2.0)*(6.83931e-02-1.15261e-03*x-8.16098e-05*x*x)",20.,60.);
// Just fit the histogram (1 parameter fit)
asl[0]->Fit("funl");
then root (minuit) will return the error the dilution parameter and the
fit uncertainty (its error bar).
-- t
________________________________________________________________________________
Tancredi Botto, phone: +1-617-253-9204 mobile: +1-978-490-4124
research scientist MIT/Bates, 21 Manning Av Middleton MA, 01949
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
On Mon, 8 Sep 2003, Hauke Kolster wrote:
> Vitaly,
> Good point about the error bars. We are looking into a 10% change in
> target
> polarization when the atomic fraction changes by 20% (60% down to 40%).
> The
> errors of the asymmetry measurements should be in the same range to be
> able
> to find a reasonable correlation to the atomic fraction.
>
> Hauke
>
>
> On Monday, September 8, 2003, at 09:49 AM, vitaliy ziskin wrote:
>
> > I think that to find a correlation between the AF and the dilution it
> > is best to use a chi squared minimization that is used in
> > show_ep_asym2.C Right now we are trying to see this correlation by
> > looking at the asymmetry in the lowest Q^2 point. This is not as
> > sensative let alone it has large error bar. The reson why TB cannot
> > make any conclusion about the AF dependence is due to large error
> > bars. If we use full acceptance (24-50) we reduce the error bar in
> > the dilution. I'm working on calculating the error bar for the
> > dilution from knowing the chi square.
>
This archive was generated by hypermail 2.1.2 : Mon Feb 24 2014 - 14:07:29 EST