Table of Contents
Name | Direction | Type | Default | Description |
---|---|---|---|---|
InputWorkspace | Input | MatrixWorkspace | Mandatory | Workspace to match the spectra between |
OutputWorkspace | Output | MatrixWorkspace | Mandatory | Workspace with the spectra matched |
ReferenceSpectrum | Input | number | 1 | Spectrum to match other spectra to |
CalculateOffset | Input | boolean | True | Calculate vertical shift |
CalculateScale | Input | boolean | True | Calculate scale factor |
Offset | Output | dbl list | Additive factor from matching | |
Scale | Output | dbl list | Multiplicitive factor from matching | |
ChiSq | Output | dbl list | Unweighted ChiSq between the spectrum and the reference. NaN means that the spectrum was not matched |
Calculate the factor(s) that best match the individual spectra to a reference spectrum using Gauss-Markov process. This is the same algorithm used by the “blend” step of PDFgetN. The data undergoes the linear transformation
such that the distance between and the
transformed spectrum is minimized. Only the portions of the spectra
where the spectra overlap with the same x-values and the uncertainties
are greater than zero are considered.
Note
Gauss-Markov does not take into account varying uncertainties when calculating the scale and offset.
import numpy as np
x = np.arange(100, dtype=np.float)
x = np.tile(x, 2)
y = np.arange(100, dtype=np.float)
y = np.tile(y, 2)
y[100:200] += 10
dy = np.zeros(y.size, dtype=np.float) + 1.
CreateWorkspace(OutputWorkspace='MatchSpectra_input',
DataX=x,
DataY=y,
DataE=dy,
NSpec=2)
_, offset, scale, chisq = MatchSpectra(InputWorkspace='MatchSpectra_input',
OutputWorkspace='MatchSpectra_output',
ReferenceSpectrum=2,
CalculateOffset=True,
CalculateScale=False)
for i in range(2):
print('spectra {}: {:.1f} * y + {:.1f}, chisq={:.1f}'.format(i+1, scale[i], offset[i], chisq[i]))
Output:
spectra 1: 1.0 * y + 10.0, chisq=0.0
spectra 2: 1.0 * y + 0.0, chisq=0.0
Categories: Algorithm Index | Diffraction\Reduction
Python: MatchSpectra.py (last modified: 2018-10-05)