Selecting Curve Fit Band Types and Limits
The Curve Fit dialog provides a means for selecting the band type, limits, and
number of iterations used for a curve fit. You can also save and load curve
fit settings from this dialog. Use What's This help for information on the individual fields.
Select Curve Fit from the Data menu to begin curve fitting.
Band Types
Three band (curve) types are provided for curve fitting: Gaussian,
Gaussian-Lorentzian, and Asymmetric Gaussian-Lorentzian. The Guassian band type is the
least complex, using only three parameters per band. This allows for a faster
fit and makes manual parameter adjustments easier. Asymmetric
Guassian-Lorentzian is the most complex, with six parameters.
The type of curve (band) type you use depends on the shape of your data peaks.
Starting with the correct band type for your data greatly enhances the
chances of a good fit. In general, use the least complex type that will produce a
good fit.
You can change band types at any time in the curve fit procedures, even after
you’ve added bands, and your data will be retained. However, note that if you
change from a more complex to less complex band type, the information represented
by any unused parameters will be lost.
Limits
The default limits determine the range within which the computer adjusts a
given parameter during a fit routine. The limits reflect the estimate of a
particular band +/- the value entered. For example, given a position estimate of
200eV, a default limit of +/- 1eV for position allows the position to be adjusted
only within the range of 199eV to 201eV. Note that some limits may be grayed
out if they do not pertain to a given band type.
“Loose” limits (limits that provide more than a sufficient range for a particular
estimate) gives the program more room for adjusting the estimates of each band but
may result in estimates that fall outside the range of expected outcomes. “Tight” limits (limits that greatly constrain an estimate) may keep estimates closer
to their expected values but may not allow enough room for adjustment to
perform a good curve fit.
Important: The limits for each parameter adjust with the estimate when the estimate is
adjusted manually, and after a curve fit routine is performed. This is why a
second command to perform a curve fit may result in a different (and possible
improved) fit after the curve fit routine has completed. Since the limits track
the estimates, they are “loosened” after each fit. Click here for information on how to lock an estimate or fit the limits asymmetrically.
Maximum Iterations
The maximum number of iterations instructs the fit routine to attempt to fit a
given set of bands up to the given iterations value. For each iteration, the
fit routine performs a non-linear regression algorithm on the curve fit band
parameters. If the fit does not improve within a specified range during an
iteration and the routine has reached convergence, it stops before the maximum
number of iterations is reached. A higher value for the maximum number of
iterations means a longer time to fit but may result in a better fit. Note that the
maximum iterations has no bearing on the way the curve fit algorithm performs,
only on how many iterations it can go before it stops.
Saving and Loading Curve Fit Information
Curve fit information is saved with your acquisition files. Additionally, you
can save curve fit files separately from your data. This enables you to store
the most commonly used curve fit settings and apply them to different
acquisitions.
Click the Save button to save the current curve fit information. All curve
fit limits, bands, and band parameters are saved. If you only want to store
default limit information, delete any bands before saving the file. To load,
simply click the Load button. Any current curve fit information is replaced.
When loading a curve fit file that contains some bands, note that the
intensities of a set of curve fit bands may not fall within the range of data for that
acquisition. For example, you may have saved a set of bands for data that
averaged 5,000 counts/seconds, but the current data you are fitting may have a
range of 10,000 counts/seconds. The software will automatically try and compensate
based on the highest peak. However, if the curve shape of the new data is
sufficiantly different from that which the fit was originally created on, you may
need to adjust the band intensities somewhat to compensate.
Adding and Removing Curve Fit Bands
Additional Curve Fitting Topics