|
Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.mso.anu.edu.au/~mwhite/pynifs-doc/pynifs.velfit-class.html
Дата изменения: Unknown Дата индексирования: Fri Feb 28 01:50:38 2014 Кодировка: Поисковые слова: annular solar eclipse |
| Home | Trees | Indices | Help |
|---|
|
|
A class to hold information by fitting multi-component Gaussians to a
NIFS data cube.
Attributes:
restwavl - The rest wavelength used to determine line velocities.
vmin - The minimum velocity value used for generating the spectrum to
fit to.
vmax - The maximum velocity value used for generating the spectrum to
fit to.
scidata - The scidata from the fitscube, trimmed down so that it only
includes the data between vmin and vmax
vars - The variance array from the fitscube, trimmed down to the
velocity range of interest
dq - The DQ pixel quality array from the fitscube, trimmed down to
the velocity range of interest
uncertest - A 2D masked array, of the same shape as the last two axes of
scidata, which contains the uncertainty estimate for each
spaxel. This uncertainty estimate is derived by determining the
1-sigma standard deviation of the science data in the original
data cube, just outside the region of interest.
xpts - The x-coordinates from the nifscube
ypts - The y-coordinates from the nifscube
vpts - The z-coordinates from the nifscube, re-expressed as km/s
velocities.
fitno - A 2D masked array, of the same shape as the last two axes of
scidata, which holds the optimal number of components fit to
each spaxel. This has been determined by a statistical F-test.
fitorig - A 2D masked array, of the same shape as fitno. This is a copy
of fitno, recorded BEFORE any masking if performed (i.e. by
maskFits.)
dof - A 2D masked array, of the same shape as the spatial axes of
scidata, holding lists of length fitmax. Holds the degrees of
freedom (dof) for each spaxel when fit with a number of
Gaussian components, up to (and including) fitmax components.
For example, the dof for a one-Gaussian fit to a given spaxel
will be given by dof[j,i][0].
chisq - As for dof, however, it holds the chi-squared value (NOT
REDUCED) for each fit to the spaxel.
fits - A 2D masked array, of the same shape as scidata, holding the
fit information for each spaxel. The elements of fits take the
form of a list. The first element will be the fit parameters
corresponding to a one-Gaussian fit, the next the parameters
for a two-Gaussian fit, and so on.
perror - As for fits, but holding the formal perror covariance matrix
values for each fit.
ftestlim - The maximum value allowed for an F-test ratio to be before the
P-value is taken. Defaults to 1.4.
ftestp - The maximum P-value permitted for two fits to be considered
statistically different under the F-test. Defaults to 0.05 (5%).
Methods:
fitVelSpax - Computes a multi-dimensional Gaussian fit to one particular
spaxel.
plotSpaxSpec - Display the data, and fit components, for a particular
spaxel.
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|||
Compute a multi-component Gaussian fit to spaxel x=i, y=j in the velfits
instance. Modifies the velfits instance in place.
If params are not provided, the function uses the following algorithm
to produce initial guesses:
- Make a copy of the spaxel data
- Find the highest data value within the range defined by vlim, use to
give guess velocity shift and amplitude
- Compute estimate for line width
- Find nearest point to peak <= half the guess amplitude, compute
where the line between the that point, and the next point
towards the peak, is = half the guess amplitude
- If the next point towards the peak is masked, use the line
between the peak and the half-amplitude point instead
- Compute the maximum allowed width for this fitted line, based on the
guess velocity centroid
- If guess width is above this value, adjust guess to 5.0 km/s
below max allowed value
- If guess width is below wmin, adjust guess to 5.0 km/s above max
allowed value
- Subtract the guessed Gaussian profile from copy of spaxel data
- Repeat the procedure until there are enough guess parameters to fit
comps components
Inputs:
velfits - The velfits instance to be acted upon.
i - The x-index of the spaxel to be fitted.
j - The y-index of the spaxel to be fitted.
comps - The number of Gaussian components to fit to the spaxel
spectrum. Defaults to 1.
fixno - Boolean value denoting whether to force the spectrum to be
fit with comps components (True), or to incrementally try a
number of components from 1 to comps, stopping when the
limiting reduced chi-squared is acheived (False). Defaults
to False.
ignoreamp - Boolean value specifying whether to consider the nifscube's
stated amplitude limit when fitting. Defaults to False (any
fits not conforming to the limit will be considered bad).
quiet - Boolean value denoting whether or not the MPFIT fitting
routine should be quiet or not. Defaults to False.
params - Fitting parameters to be passed to the fitting routine.
Intended for use when the automatic paramter finder produces
unacceptable results. Requires fixno = True.
Takes the form of a list composed of
[background, amplitude, shift, width, amplitude....].
Defaults to None (at which point the
automatic first-guess finding will be used.)
vlim - A two-element list giving the minimum and maximum values
allowed for ALL component centroids in the fits, in units
of km/s. Defaults to None (where the minimum and maximum
velocities present in the fitvels instance will be used as
the limits.
wmin - The minimum allowed line sigma to be fit. Note this does NOT
represent a FWHM value. Defaults to 10.0
wpercent - The percentage of total line intensity allowed to be on the
opposite side of v=0 to the line peak. Used to calculate an
upper limit on the fitted line width. Must be a float
between 0.0 and 1.0. Defaults to 0.10.
wperspace - Expected percentage error on the initial guess for the line
centroid velocity. Used in the calculation of maximum
allowed fitted line width. Must be a float between 0.0 and
1.0. Defaults to 0.25.
wmaxcut - Lower limit on maximum allowed fitted line width. Avoids the
initial guess routine from trying to fit peaks of noise.
Defaults to 50.0 km/s.
alim - A two-element list giving the minimum and maximum values
allowed for the amplitude of ALL fit components, in units
km/s. Defaults to None (at which point, the amplitudes
will only be limited to be positive).
fixbg - Boolean value, specifying if the background height of the
fits should be forced to 0.0 (True) or not. Defaults to
False.
snthresh - Signal-to-noise threshold - if spaxel is below this
threshold, fits will still be made, but fitno[j,i] will then
be set to 0 afterwards. Masking calculated by the
maskFits function.
Returns:
Nil. The velfits instance is updated in place. The fit information for
the i-th, j-th pixel will be updated as follows:
If fixno = False, the fit information for the first multi-component
fit that satisfies the rchisqlim for this velfits instance will be
recorded. If all fits, up to and including the comps-component fit,
do not satisfy this limit, no fit info (and empty list) will be
recorded.
If fixno = True, then the fit will be made blindly and recorded,
with no regard to the reduced chi-squared limit.
|
'Mask' spaxel if signal-to-noise is under the specified value. By
'mask', the no. of fits recorded for that spaxel will be set to 0.
Signal-to-noise is determined as the ratio of the peak flux value
recorded in the spaxel to the corresponding error of that data point.
Inputs:
i, j - Spaxel x- and y- indices, respectively.
snthresh - The S/N threshold value. Any spaxels with an S/N less than this
value will be 'masked'. Defaults to 7.5.
Returns:
Nil. fitvels instance is updated in-situ.
|
Mask a particular pixel in the velfit instance. Inputs: i, j, k - The pixel's x-, y- and spectral indices respectively. Returns: Nil. The velfit instance is updated in-situ. |
Apply maskFits to every spaxel in the velfit instances. Inputs: snthresh - The S/N threshold for masking. Defaults to 5. Returns: Nil. fitvels instance updated in-situ. |
|
|