Документ взят из кэша поисковой машины. Адрес оригинального документа : http://hea-www.harvard.edu/PINTofALE/doc/FITLINES.howto Дата изменения: Tue May 16 01:42:55 2006 Дата индексирования: Tue Oct 2 01:09:10 2012 Кодировка: Поисковые слова: http www.astronomy.ru forum index.php topic 4644.0.html

How to use FITLINES
===================

FITLINES is an interactive widget-based function to fit 1-dimensional
profiles to features in 1-D spectral data, Y[X]. In principle, it
only works with 3-parameter curve families (Gaussians, Lorentzians),
but in practice curves with extra parameters may be fit (Modified
Lorentzians, Rotationally Broadened Line) provided that the extra
parameters are held fixed.

Currently, the following functions are encoded via MK_3MODEL():

Function TYPE definitions for MK_3MODEL()

Gaussian G; GAUSSIAN; etc.
Lorentzian L; LORENTZIAN; etc.
Beta-profile B; BETA; BETA=; BETA:; BETA ; etc.
Asymmetric Slanted SL; SLANT; SLANT=; SLANT=,; etc.
Beta-profile
Rotation Broadened R; ROT; ROT=; etc.
Gaussian
Broken Power-Law PW; PW@; PW=; etc.
Identity I; Ident; etc.
Sinusoid SI; Sin=; etc.

FITLINES must be called as a function, and returns a structure containing
the best-fit positions, fluxes, widths, associated errors, etc.

To see the complete calling sequence, type
help,fitlines()
at the "IDL>" prompt. The output lists all the parameters and keywords
accepted by FITLINES. Do not panic: in the majority of the cases, only
a few of them have to be specified. For a description on what each of
the variables mean, see
http://hea-www.harvard.edu/PINTofALE/PoA.html#FITLINES

An example walk-through for using FITLINES is given below.
This is by no means to be considered comprehensive, but
does exercise some of the major features of the GUI.

-- Start IDL

-- initialize the PINTofALE environment

.run initale

-- restore a spectrum for fitting line profiles to (we will
fit beta-profiles to the Si XIII triplet at 6.7 AA):

restore,!ARDB+'/demo_v711Tau.save',/verbose

plot,LAM_M1P_V711TAU,SPC_M1P_V711TAU,psym=10,xr=[6.5,7.0],/xs,$
xtitle='Wavelength ['+string(byte(197))+']',ytitle='Counts/bin',$
title='HR 1099: HETG/MEG +1',yr=[0,400],/ys

-- initialize the input to FITLINES

;Specify that 3 model components are needed by declaring a 3-element
;array for the locations. the locations need only be approximate.
pos=fltarr(3)+6.5

NOTE: it is possible to start off with fewer components; new ones
may be added to the mix using the GUI.

type='beta 1.8' ;to fit beta-profiles of index 1.8
wdt=0.05 ;set the initial values of the widths to 0.05

-- start up the GUI

fstr=fitlines(LAM_M1P_V711TAU,SPC_M1P_V711TAU,ysig=SPE_M1P_V711TAU,$
pos=pos,perrp=perrp,perrm=perrm,perrc=perrc,$
flx=flx,ferrp=ferrp,ferrm=ferrm,ferrc=ferrc,$
wdt=wdt,werrp=werrp,werrm=werrm,werrc=werrc,$
thaw=thaw,type=type,epithet=epithet,ties=ties,$
conlev=conlev,consig=consig,/dumb,verbose=!verbose)

NOTE: the /DUMB keyword ensures that FIT_LEVMAR does not stop every
so often to ask for directions. For those of you raised on XSPEC,
that is like setting query to yes while fitting.

-- display the full spectrum

[PULL DOWN] VIEW -> SHOW

-- restrict X-axis range

[PULL DOWN] XRANGE -> Keyboard
[TYPE @ PROMPT] 6.5 7
[PULL DOWN] VIEW -> SHOW

where "" represents a carriage-return.

To also set the Y-axis range to something more useful,

[PULL DOWN] YRANGE -> Current
[PULL DOWN] VIEW -> SHOW

-- the initial guesses for the line locations are clearly very bad.
reset them to more reasonable values at the parameter editing
interface at center-left of the GUI.

[TYPE IN BOX "Component"] 0
[TYPE IN BOX "Position"] 6.65
[TYPE IN BOX "Label"] Si13 r
[TYPE IN BOX "Component"] 1
[TYPE IN BOX "Position"] 6.70
[TYPE IN BOX "Label"] Si13 i
[TYPE IN BOX "Component"] 2
[TYPE IN BOX "Position"] 6.75
[TYPE IN BOX "Label"] Si13 f

-- it is better to let the program figure out the best guesses
for the initial values of the fluxes, so click on the buttons
corresponding to the fluxes (parameters A2, A5, and A8) at the
left of the center-right descriptor window, and then renormalize
the fluxes

[LEFT CLICK ON BUTTON FOR A2]
[LEFT CLICK ON BUTTON FOR A5]
[LEFT CLICK ON BUTTON FOR A8]
[LEFT CLICK] RENORM

the model profiles at this point look thus:

[PULL DOWN] VIEW -> SHOW

-- there is also a fairly large continuum at this wavelength range.
the continuum is best set "by hand".

[PULL DOWN] CONTINUUM -> Piecewise
[LEFT CLICK ON PLOT, AT APPROX (6.9,95)]
(multiple clicks will allow a more complex definition)
[LEFT CLICK IN PLOT WINDOW, BELOW X-AXIS]
(get rid of default Poisson error on continuum estimation)
[RIGHT CLICK WITHIN PLOT]
(exit)
[PULL DOWN] CONTINUUM -> ACCEPT
(to adopt the newly defined continuum) (VERY IMPORTANT!!!)

[PULL DOWN] VIEW -> SHOW
[LEFT CLICK] RENORM
[PULL DOWN] VIEW -> SHOW

-- set up constraints on the parameters (these are essentially
any legal IDL statement that modify the parameter values)

these ensure that the widths of all the fitted lines are identical:
[TYPE IN BOX "ADD"] A4 = A1
[TYPE IN BOX "ADD"] A7 = A1

these ensure that all fluxes are +ve:
[TYPE IN BOX "ADD"] A2 = A2 > 0
[TYPE IN BOX "ADD"] A5 = A5 > 0
[TYPE IN BOX "ADD"] A8 = A8 > 0

far more complicated constraints may also be set. for example,
the following ensures that the flux of the intercombination line
lies between a fifth to half that of the forbidden line
[TYPE IN BOX "ADD"] A5 = (A5 < (A8/2.) > (0.2*A8))

as another example, the following forces the location of the
intercombination line to lie between that of the resonance and
the forbidden lines
[TYPE IN BOX "ADD"] A3 = (A3 > A0) < A6

-- Now thaw all the relevant parameters

[LEFT CLICK ON BUTTON FOR A0]
[LEFT CLICK ON BUTTON FOR A1]
[LEFT CLICK ON BUTTON FOR A3]
[LEFT CLICK ON BUTTON FOR A6]

Note that A4 and A7 must be left frozen, because they are
to have values identical to A1. However, A3, A5, A6, and
A7 are thawed, because even if strongly constrained, they
must still be fit.

-- Fit

[LEFT CLICK] FIT
[PULL DOWN] VIEW -> SHOW

which results in a reasonable fit (reduced chisq ~ 1.1)

the widths may be -ve, which actually does not matter except for
aesthetic reasons. if -ve, reset them to +ve values using the
parameter editing interface.

-- reduce the range over which the fit is carried out. let the
program guess at an appropriate range, which it does by looking
at the values of all the thawed positions and widths.

[PULL DOWN] XRANGE -> Guess

-- and fit again

[PULL DOWN] VIEW -> SHOW
[LEFT CLICK] FIT
[PULL DOWN] VIEW -> SHOW

-- save the fit parameters internally

[LEFT CLICK] SAVE

-- dump the current state to disk, for leisurely access via FITLINES_UNDUMP

[LEFT CLICK] DUMP
[TYPE @ PROMPT] /tmp/fitlines.save

-- Obtain 1-sigma chisq-surface-projected errors

[LEFT CLICK] ERRORS

NOTE: the error finding subroutine is very robust, but it may still
crash (or go into a catatonic loop) when confronted with pathological
error surfaces. it is always possible to get back to the GUI in such
cases by typing RETURN at the ERORS subroutine level.

-- fit to a new line, say the one at 6.2 AA

[PULL DOWN] XRANGE -> RESET
[PULL DOWN] XRANGE -> Keyboard
[TYPE @ PROMPT] 5.5 7.0
[PULL DOWN] YRANGE -> Current

Note the sharp drop at 6.5 AA, which is due to the fact that the
continuum was defined only in the range 6.5-7.0 AA. Choose a small
interval in which to fit the new line:

[PULL DOWN] XRANGE -> Cursor
this will put you in PICKRANGE(), which is described elsewhere.
All that matters here is to hold down the left button and drag
it over the x-interval of interest.
[LEFT DRAG, say from (6.1,100) to (6.3,100)]
[RIGHT CLICK to exit PICKRANGE()]

define a new model component
[PULL DOWN] MODEL -> ADD Component
hold down the left button and drag from the peak of the line
to a point that denotes the width
[DRAG, say from (6.19,350) to (6.2,150)]
[CLICK to exit model definition]

in order to fit only the new component to the selected line feature,
all the other model components must be frozen
[PULL DOWN] MODEL -> Freeze Unimportant
[LEFT CLICK] RENORM

the continuum must be redefined
[PULL DOWN] CONTINUUM -> Piecewise
to avoid using Poisson errors on the defined continuum,
[LEFT CLICK BELOW X-AXIS]
to adjust the continuum level, hold down the left button and drag
to new level
[LEFT DRAG, from say (6.2,95) to (6.2,90)]
[RIGHT CLICK to exit]
[PULL DOWN] CONTINUUM -> ACCEPT
[LEFT CLICK] RENORM
[PULL DOWN] VIEW -> SHOW

[LEFT CLICK] FIT
[PULL DOWN] VIEW -> SHOW

which is again a reasonable fit, with a reduced chisq ~ 1.4

[LEFT CLICK] ERRORS

-- to quit

[LEFT CLICK] QUIT

-- all the useful quantities are in the output,
IDL> help,fstr,/str
** Structure <39a2d0>, 21 tags, length=196952, refs=1:
POS FLOAT Array[4]
PERRP FLOAT Array[4]
PERRM FLOAT Array[4]
PERRC FLOAT Array[4]
FLX FLOAT Array[4]
FERRP FLOAT Array[4]
FERRM FLOAT Array[4]
FERRC FLOAT Array[4]
WDT FLOAT Array[4]
WERRP FLOAT Array[4]
WERRM FLOAT Array[4]
WERRC FLOAT Array[4]
THAW INT Array[12]
TYPE STRING Array[4]
TIES STRING Array[7]
EPITHET STRING Array[4]
CONLEVX DOUBLE Array[8192]
CONLEV FLOAT Array[8192]
CONSIGX DOUBLE Array[8192]
CONSIG FLOAT Array[8192]
COMMENT STRING ' {a4 ='...
IDL> for i=0,3 do print,(fstr.pos)[i]+[-(fstr.perrm)[i],0,(fstr.perrp)[i]]
6.65619 6.65661 6.65703
6.69431 6.69571 6.69710
6.74903 6.74935 6.74967
6.18996 6.19036 6.19075
IDL> for i=0,3 do print,(fstr.wdt)[i]+[-(fstr.werrm)[i],0,(fstr.werrp)[i]]
0.0111397 0.0116787 0.0122177
0.0111397 0.0116787 0.0122177
0.0111397 0.0116787 0.0122177
0.0115940 0.0122118 0.0128578
IDL> for i=0,3 do print,(fstr.flx)[i]+[-(fstr.ferrm)[i],0,(fstr.ferrp)[i]]
1063.90 1109.42 1154.95
152.692 184.313 215.934
612.465 651.284 690.103
1155.23 1205.77 1256.64