This panel pops up once you have constructed a model and clicked on the Construct
button in the Fit Panel. It has the following components:
- Parameter entry: Each fitting parameter is given a
line with some boxes. Next to the name of the parameter is a check box. Check this
if you want this parameter varied during the fit; otherwise leave it unchecked. (Notice that "Area"
type parameters cannot be varied as they are not truly independent--they are calculated from
the amlitude and the width). Next is
the "Value" of the parameter. Enter an initial guess for the parameter; the final
value after the least-squares minimization will also appear here.
Next is the one-parameter error bar. This is an estimate of the one-standard-deviation
uncertainty in this parameter, assuming that no other parameters were varied.
Finally, there are low and high values for the "Multi-parameter error bar." These
are one-standard deviation limits on this parameter assuming that all other variable parameters
(those which you have allowed to vary)
are adjusted simultaneously.
- Apply Button: Click here to calculate the model with
these starting parameters, not allowing anything to vary. The results of your fit
will appear together with the original data in the
Line Plot Image Window.
- Fit Button: Click here to perform a least-squares fit,
using the starting parameters you entered in the Value box for each parameter, and
varying only the checked parameters.
The results of your fit
will appear together with the original data in the
Line Plot Image Window. Useful messages may appear in the
Fit Message Area of the
Fit Panel. The new values of the fitting parameters will appear
in the Value boxes, and the estimated uncertainties will appear in the boxes to the
right of these.
- Correlations Button: Click here to
bring up a new window showing the correlations between different parameters.
A correlation coeficient of 0 means that two parameters are "orthogonal"--they
have totally different effects on the shape of the curve. A correlation coefficient
of 1 means that the two parameters are totally correlated, and do exactly the same
thing to the model--this is a bad thing. Similarly, a correlation of -1 is bad.
You generally bring up this window if a fit fails due to strongly correlated parameters
and you want to figure out which ones.
- Revert Button: Click here to set all parameters back
to the values they had before the list time you performed a fit.
- Cancel: If you realize that you made a mistake in selecting files you
want to open, cancel the opening process by clicking this button.
- Reset: Set all parameters back to reasonable values such that the
function is at least likely to display properly on the plot.
- MPEB Box : Check this box to calculate multi-parameter
error bars after doing the fit. This slows things down a bit, so you may want to only
do this part for the very last fit.
- Display Submodels Box: Check this box to plot all the
individual submodels in addition to the total model.
- Display Fit Cursors: Most fit models come equipped with cursors, which allow you to graphically
change some or all of the fit parameters. If this box is unchecked, the
fit cursors are not shown. (The default is to show all cursors).
The general procedure for performing a least-squares fit is as follows:
- Open a data file in the usual way. Make a line plot. The data in the line plot
will be what you are fitting to.
- Use the
Fit Panel to construct a model function by specifying the
number of sub-functions and the type for each, and click on
Construct to bring up the
Fit Parameter Panel.
- Set each parameter to a reasonable starting value.
- Click the
Apply Button and see whether your model with these
parameters agrees approximately with the data. If it does not, go back to step 3.
- Check off the parameters you want to vary, and click on the
Fit Button. If you get unexpected results, you may want to
uncheck a few parameters, click on the Revert Button, and
try again.
- To cancel a fit in progress, click on the Cancel Button.
- Check in the
message area to see how well the fit worked, and whether there
were any errors.
- If you are happy with the results, you may want to
print the plot showing the agreement between model and data,
or save< the fitting parameters.
See Help -> Least Squares Fits for more information.
|