Refinement: dropping unnecessary peak parameters
After HyperLab found an optimum model for the fitting problem, it eliminates unnecessary (insignificant) fitting parameters. Basis of the elimination may be too small value, too small value compared to its uncertainty etc.
Sometimes this logic does not drop parameters which are unnecessary for a good fit, and user intervention will be required – or simply two models fit well to a specific region, and the spectroscopist would choose a model which was not preferred by the automatic fitter.
The figure shows a region where the algorithm chooses a significant step for the background, but one can suspect that another model without a step would also fit well to the measured counts.
To eliminate the step parameter from the region's fitted model, right-click the region. A popup menu appears now, providing the parameters that may be eliminated. Choose Drop Background Step menu item.
HyperLab now eliminates the step, refits the region and displays the new fit.
The Chi-squared value increased (from 1.9 to 2.4), showing why the automatic fitter inserted the step previously.
But, if you check the residual, you can see that one of the most deviating points is the last one in the region – that is, the background has a changing tendency.
In order to further improve the fit, move its right boundary to leftwards by one channel. This leads to improved Chi-square (2.1) and an acceptable fit.
Copyright © 1998-2007 by HyperLabs Software Budapest, Hungary