def aperphot(fn, timekey=None, pos=[0,0], dap=[2,4,6], mask=None, verbose=False, nanval=999, resamp=None, retfull=False, ignorenan=True):
"""Do aperture photometry on a specified file.
:INPUTS:
pos : 2-sequence
center of apertures (as if indexing: fn[pos[0], pos[1]])
dap : 3-sequence
Photometry aperture DIAMETERS:
-- target aperture (within which to sum flux)
-- inner sky aperture
-- outer sky aperture
resamp : int
Factor by which to interpolate frame before measuring photometry
(in essence, does partial-pixel aperture photometry)
Aperture masking:
If no mask is passed in, use the star's input position and
aperture diameters to create binary pixel masks for stellar and
background photometry. If a mask is passed in, stellar and
background aperture masks are generated from where all input
mask elements equal 1 and 2, respectively.
retfull:
Also return arrays of target mask, sky mask, and frame used.
This option is a memory hog!
:OUTPUTS:
:class:`phot` object.
:EXAMPLE:
::
import astropy.io.fits
from astropy import wcs
import numpy as np
from phot import aperphot
img='62_z_CDFs_goods_stamp_img.fits' #path to the image
RA = 52.9898239
DEC = -27.7143114
hdulist = astropy.io.fits.open(img)
w = wcs.WCS(hdulist['PRIMARY'].header)
world = np.array([[RA, DEC]])
pix = w.wcs_world2pix(world,1) # Pixel coordinates of (RA, DEC)
print "Pixel Coordinates: ", pix[0,0], pix[0,1]
#call aperture function
observation=aperphot(img, timekey=None, pos=[pix[0,0], pix[0,1]], dap=[4,8,12], resamp=2, retfull=False)
# Print outputs
print "Aperture flux:", observation.phot
print "Background: ", observation.bg
:REQUIREMENTS:
scipy.interpolate, pyfits, numpy...
"""
# 2009-09-14 10:49 IJC: Created
# 2010-01-15 14:20 IJC: Added numpy "_string" check
# 2011-12-29 12:01 IJMC: Added peak pixel values to photometry report.
# 2012-01-25 11:26 IJMC: Adding "resamp" option -- thanks to
# K. Stevenson and J. Harrington of UCF for
# the suggestion.
# 2012-02-26 11:53 IJMC: Now return 'ntarg' and 'nsky' -- number of pixels used.
# 2012-06-07 08:27 IJMC: 'peak' values are now corrected for the
# resampling factor.
# 2012-07-03 10:35 IJMC: Fixed a key bug: frames were not
# correctly background-subtracted when
# applying partial-pixel resampling.
# 2012-10-19 13:41 IJMC: Documented 'retfull' option; changed default.
# 2013-03-20 09:21 IJMC: More error-checking for saving header
# keywords. Thanks to A. Weigel @
# ETH-Zurich for catching this!
# 2014-05-07 16:05: added "ephot" by J. Xavier Prochaska, UCSC
# 2014-08-28 09:32 IJMC: Added better checking for user-input
# masks and resamp>1. Corrected x/y
# indexing for non-square inputs arrays.
# 2014-09-06 14:41 IJMC: Added better checking for dap & pos.
# 2014-09-08 10:10 IJMC: Fix masking & nan-handling.
from numpy import meshgrid, median,isfinite,sort,ndarray,string_
import numpy as np
try:
from astropy.io import fits as pyfits
except:
import pyfits
from analysis import fixval
from os import path
from scipy import interpolate
thisobs = phot()
if pos is None:
pos = 0, 0
if dap is None:
dap = 1, 2, 3
x0, y0 = pos
dap_targ, dap_skyinner, dap_skyouter = dap
if resamp is None or resamp<1:
resamp = 1
else:
resamp = float(resamp)
# Determine size:
if isinstance(fn,str):
nx = pyfits.getval(fn, 'NAXIS2')
ny = pyfits.getval(fn, 'NAXIS1')
elif isinstance(fn,ndarray):
nx,ny = fn.shape
nx0, ny0 = nx, ny
nx = ((nx - 1)*resamp + 1.) # Avoid resampling at pixel locations
ny = ((ny - 1)*resamp + 1.) # outside the original boundaries.
# Generate or load masks:
if mask is None:
xx,yy = meshgrid(np.arange(ny)/resamp, np.arange(nx)/resamp)
mask_targ = makemask(xx, yy, (y0, x0, dap_targ))
mask_s1 = makemask(xx, yy, (y0,x0, dap_skyinner))
mask_s2 = makemask(xx, yy, (y0,x0, dap_skyouter))
mask_sky = mask_s2 - mask_s1
userInputMask = False
#pdb.set_trace()
else:
mask_targ = mask==1
mask_sky = mask==2
userInputMask = True
# Load data frame:
thisobs = phot()
if pos is None:
pos = 0, 0
if dap is None:
dap = 1, 2, 3
if isinstance(fn,ndarray):
frame = fn
elif isinstance(fn, str) or isinstance(fn,string_):
if not path.isfile(fn):
print "file %s not found! exiting..." % fn
return thisobs
frame = pyfits.getdata(fn)
fixval(frame, nanval)
# Resample data frame
badval = -12345.6789
if resamp>1:
frame0 = frame.copy()
mask_good0 = np.isfinite(frame0)
frame0[True-mask_good0] = badval
xx0 = range(nx0)
yy0 = range(ny0)
x1,y1 = np.arange(nx)/resamp, np.arange(ny)/resamp
rectspline = interpolate.fitpack2.RectBivariateSpline(xx0, yy0, frame0, kx=1, ky=1, s=0)
frame = rectspline(x1, y1)
if userInputMask and \
((frame.shape<>mask_targ.shape) or (frame.shape<>mask_sky.shape)):
print "User-entered mask did not have the correct size for the resampling"
print " input used (resamp=%1.3f). Beware!" % resamp
stop
if ignorenan:
mask_good = np.isfinite(frame)
mask_targ = mask_targ * mask_good
mask_sky = mask_sky * mask_good
#from pylab import *
#pdb.set_trace()
# Measure background and aperture photometry
thisbg, thisebg = estbg(frame, mask=mask_sky, plotalot=verbose, rout=[3,99], verbose=verbose)
if not np.isfinite(thisbg):
thisbg = 0.
thisphot = ((frame - thisbg)[mask_targ]).sum() /resamp/resamp
peak = frame.max()
peak_targ = frame[mask_targ].max()
peak_annulus = frame[mask_sky].max()
thisobs.bg=thisbg
thisobs.ebg=thisebg
thisobs.bgstr='phot.estbg: SDOM on bg histogram mean & dispersion after outlier rejection'
thisobs.phot=thisphot
thisobs.photstr='by-hand background-subtracted aperture photometry'
thisobs.ntarg = mask_targ.sum()/resamp/resamp
thisobs.nsky = mask_sky.sum()/resamp/resamp
thisobs.peak = peak
thisobs.peak_targ = peak_targ
thisobs.peak_annulus = peak_annulus
thisobs.peakstr = 'peak pixel value in frame'
thisobs.peak_targstr = 'peak pixel value in target aperture'
thisobs.peak_annulusstr = 'peak pixel value in sky annulus'
thisobs.position = pos
thisobs.positionstr = 'user-specified, zero-indexed pixel coordinates.'
if isinstance(fn, str):
header = pyfits.getheader(fn)
if not timekey is None:
if timekey in header:
thisobs.time=header['timekey']
thisobs.timestr='heliocentric modified julian date'
if 'object' in header: thisobs.object = header['object']
if 'exptime' in header: thisobs.exptime = header['exptime']
thisobs.aper = dap
thisobs.aperstr = 'target, inner, outer aperture diameters, in pixels.'
thisobs.filename=fn
thisobs.resamp = resamp
# Simple stats :: JXP 2014 May 6
var = thisphot + np.sqrt(thisobs.nsky)*thisobs.bg
thisobs.ephot = np.sqrt(var)
if retfull:
thisobs.mask_targ = mask_targ
thisobs.mask_sky = mask_sky
thisobs.frame = frame
if verbose>0:
from pylab import figure, colorbar
from nsdata import imshow
figure(); imshow(frame*mask_targ); colorbar()
figure(); imshow(frame*mask_sky); colorbar()
return thisobs
def aperphot(fn, timekey=None, pos=[0,0], dap=[2,4,6], mask=None, verbose=False, nanval=999, resamp=None, retfull=False, ignorenan=True):
"""Do aperture photometry on a specified file.
:INPUTS:
pos : 2-sequence
center of apertures (as if indexing: fn[pos[0], pos[1]])
dap : 3-sequence
Photometry aperture DIAMETERS:
-- target aperture (within which to sum flux)
-- inner sky aperture
-- outer sky aperture
resamp : int
Factor by which to interpolate frame before measuring photometry
(in essence, does partial-pixel aperture photometry)
Aperture masking:
If no mask is passed in, use the star's input position and
aperture diameters to create binary pixel masks for stellar and
background photometry. If a mask is passed in, stellar and
background aperture masks are generated from where all input
mask elements equal 1 and 2, respectively.
retfull:
Also return arrays of target mask, sky mask, and frame used.
This option is a memory hog!
:OUTPUTS:
:class:`phot` object.
:EXAMPLE:
::
import astropy.io.fits
from astropy import wcs
import numpy as np
from phot import aperphot
img='62_z_CDFs_goods_stamp_img.fits' #path to the image
RA = 52.9898239
DEC = -27.7143114
hdulist = astropy.io.fits.open(img)
w = wcs.WCS(hdulist['PRIMARY'].header)
world = np.array([[RA, DEC]])
pix = w.wcs_world2pix(world,1) # Pixel coordinates of (RA, DEC)
print "Pixel Coordinates: ", pix[0,0], pix[0,1]
#call aperture function
observation=aperphot(img, timekey=None, pos=[pix[0,0], pix[0,1]], dap=[4,8,12], resamp=2, retfull=False)
# Print outputs
print "Aperture flux:", observation.phot
print "Background: ", observation.bg
:REQUIREMENTS:
scipy.interpolate, pyfits, numpy...
"""
# 2009-09-14 10:49 IJC: Created
# 2010-01-15 14:20 IJC: Added numpy "_string" check
# 2011-12-29 12:01 IJMC: Added peak pixel values to photometry report.
# 2012-01-25 11:26 IJMC: Adding "resamp" option -- thanks to
# K. Stevenson and J. Harrington of UCF for
# the suggestion.
# 2012-02-26 11:53 IJMC: Now return 'ntarg' and 'nsky' -- number of pixels used.
# 2012-06-07 08:27 IJMC: 'peak' values are now corrected for the
# resampling factor.
# 2012-07-03 10:35 IJMC: Fixed a key bug: frames were not
# correctly background-subtracted when
# applying partial-pixel resampling.
# 2012-10-19 13:41 IJMC: Documented 'retfull' option; changed default.
# 2013-03-20 09:21 IJMC: More error-checking for saving header
# keywords. Thanks to A. Weigel @
# ETH-Zurich for catching this!
# 2014-05-07 16:05: added "ephot" by J. Xavier Prochaska, UCSC
# 2014-08-28 09:32 IJMC: Added better checking for user-input
# masks and resamp>1. Corrected x/y
# indexing for non-square inputs arrays.
# 2014-09-06 14:41 IJMC: Added better checking for dap & pos.
# 2014-09-08 10:10 IJMC: Fix masking & nan-handling.
from numpy import meshgrid, median,isfinite,sort,ndarray,string_
import numpy as np
try:
from astropy.io import fits as pyfits
except:
import pyfits
from analysis import fixval
from os import path
from scipy import interpolate
thisobs = phot()
if pos is None:
pos = 0, 0
if dap is None:
dap = 1, 2, 3
x0, y0 = pos
dap_targ, dap_skyinner, dap_skyouter = dap
if resamp is None or resamp<1:
resamp = 1
else:
resamp = float(resamp)
# Determine size:
if isinstance(fn,str):
nx = pyfits.getval(fn, 'NAXIS2')
ny = pyfits.getval(fn, 'NAXIS1')
elif isinstance(fn,ndarray):
nx,ny = fn.shape
nx0, ny0 = nx, ny
nx = ((nx - 1)*resamp + 1.) # Avoid resampling at pixel locations
ny = ((ny - 1)*resamp + 1.) # outside the original boundaries.
# Generate or load masks:
if mask==None:
xx,yy = meshgrid(np.arange(ny)/resamp, np.arange(nx)/resamp)
mask_targ = makemask(xx, yy, (y0, x0, dap_targ))
mask_s1 = makemask(xx, yy, (y0,x0, dap_skyinner))
mask_s2 = makemask(xx, yy, (y0,x0, dap_skyouter))
mask_sky = mask_s2 - mask_s1
userInputMask = False
#pdb.set_trace()
else:
mask_targ = mask==1
mask_sky = mask==2
userInputMask = True
# Load data frame:
thisobs = phot()
if pos is None:
pos = 0, 0
if dap is None:
dap = 1, 2, 3
if isinstance(fn,ndarray):
frame = fn
elif isinstance(fn, str) or isinstance(fn,string_):
if not path.isfile(fn):
print "file %s not found! exiting..." % fn
return thisobs
frame = pyfits.getdata(fn)
fixval(frame, nanval)
# Resample data frame
badval = -12345.6789
if resamp>1:
frame0 = frame.copy()
mask_good0 = np.isfinite(frame0)
frame0[True-mask_good0] = badval
xx0 = range(nx0)
yy0 = range(ny0)
x1,y1 = np.arange(nx)/resamp, np.arange(ny)/resamp
rectspline = interpolate.fitpack2.RectBivariateSpline(xx0, yy0, frame0, kx=1, ky=1, s=0)
frame = rectspline(x1, y1)
if userInputMask and \
((frame.shape<>mask_targ.shape) or (frame.shape<>mask_sky.shape)):
print "User-entered mask did not have the correct size for the resampling"
print " input used (resamp=%1.3f). Beware!" % resamp
stop
if ignorenan:
mask_good = np.isfinite(frame)
mask_targ = mask_targ * mask_good
mask_sky = mask_sky * mask_good
#from pylab import *
#pdb.set_trace()
# Measure background and aperture photometry
thisbg, thisebg = estbg(frame, mask=mask_sky, plotalot=verbose, rout=[3,99], verbose=verbose)
if not np.isfinite(thisbg):
thisbg = 0.
thisphot = ((frame - thisbg)[mask_targ]).sum() /resamp/resamp
peak = frame.max()
peak_targ = frame[mask_targ].max()
peak_annulus = frame[mask_sky].max()
thisobs.bg=thisbg
thisobs.ebg=thisebg
thisobs.bgstr='phot.estbg: SDOM on bg histogram mean & dispersion after outlier rejection'
thisobs.phot=thisphot
thisobs.photstr='by-hand background-subtracted aperture photometry'
thisobs.ntarg = mask_targ.sum()/resamp/resamp
thisobs.nsky = mask_sky.sum()/resamp/resamp
thisobs.peak = peak
thisobs.peak_targ = peak_targ
thisobs.peak_annulus = peak_annulus
thisobs.peakstr = 'peak pixel value in frame'
thisobs.peak_targstr = 'peak pixel value in target aperture'
thisobs.peak_annulusstr = 'peak pixel value in sky annulus'
thisobs.position = pos
thisobs.positionstr = 'user-specified, zero-indexed pixel coordinates.'
if isinstance(fn, str):
header = pyfits.getheader(fn)
if not timekey==None:
if timekey in header:
thisobs.time=header['timekey']
thisobs.timestr='heliocentric modified julian date'
if 'object' in header: thisobs.object = header['object']
if 'exptime' in header: thisobs.exptime = header['exptime']
thisobs.aper = dap
thisobs.aperstr = 'target, inner, outer aperture diameters, in pixels.'
thisobs.filename=fn
thisobs.resamp = resamp
# Simple stats :: JXP 2014 May 6
var = thisphot + np.sqrt(thisobs.nsky)*thisobs.bg
thisobs.ephot = np.sqrt(var)
if retfull:
thisobs.mask_targ = mask_targ
thisobs.mask_sky = mask_sky
thisobs.frame = frame
if verbose>0:
from pylab import figure, colorbar
from nsdata import imshow
figure(); imshow(frame*mask_targ); colorbar()
figure(); imshow(frame*mask_sky); colorbar()
return thisobs
def estbg(im, mask=None, bins=None, plotalot=False, rout=(3,200), badval=nan, verbose=False):
"""Estimate the background value of a masked image via histogram fitting.
INPUTS:
im -- numpy array. Input image.
OPTIONAL INPUTS:
mask -- numpy array. logical mask, False/0 in regions to ignore
bins -- sequence. edges of bins to pass to HIST
plotalot -- bool. Plot the histogram and fit.
rout -- 2-tuple of (nsigma, niter) for analysis.removeoutliers.
Set to (Inf, 0) to not cut any outliers.
badval -- value returned when things go wrong.
OUTPUT:
b, s_b -- tuple of (background, error on background) from gaussian fit.
Note that the error is analagous to the standard deviation on the mean
COMMENTS:
The fit parameters appear to be robust across a fairly wide range of bin sizes. """
# 2009-09-02 17:13 IJC: Created!
# 2009-09-04 15:07 IJC: Added RemoveOutliers option. Use only non-empty bins in fit.
# 2009-09-08 15:32 IJC: Error returned is now divided by sqrt(N) for SDOM
# 2009-11-03 00:16 IJC: Improved guess for gaussian dispersion
# 2011-05-18 11:47 IJMC: Moved (e)gaussian imports to analysis.
# 2012-01-01 21:04 IJMC: Added badval option
# 2012-08-15 17:45 IJMC: Numpy's new histogram no longer accepts 'new' keyword
# 2013-03-20 08:22 IJMC: Now works better even for small numbers
# of pixels; thanks to A. Weigel @
# ETH-Zurich for catching this!
# 2014-08-29 10:05 IJMC: Added verbosity flag.
# 2014-09-25 14:29 IJMC: Added check for no usability
from numpy import histogram, mean, median, sqrt, linspace, isfinite, ones,std
from pylab import find
from scipy import optimize
from analysis import removeoutliers, egaussian, gaussian, stdr
if plotalot:
from pylab import figure, errorbar, plot, colorbar, title, hist, mean, std
#from analysis import imshow
def gaussianChiSquared(guess, x, y, err):
return (egaussian(guess, x, y, e=err)**2).sum()
if mask is None:
mask = ones(im.shape)
dat = im.ravel()[find(mask<>0)]
if plotalot:
figure(); plot(im.ravel()); plot(dat)
print mean(dat), std(dat), rout[0]*std(dat)
print len(dat), (abs(dat-mean(dat))<(rout[0]*std(dat))).sum()
figure(); plot(dat-mean(dat));
plot([0,len(dat)], [rout[0]*std(dat),rout[0]*std(dat)],'--k')
plot([0,len(dat)], [-rout[0]*std(dat),-rout[0]*std(dat)],'--k')
dat = removeoutliers(dat, rout[0], remove='both', center='mean', niter=rout[1], verbose=plotalot)
ndat = len(dat)
if ndat==0:
if verbose>0: print "No data to work with!"
return (badval, badval)
if bins is None:
if plotalot or verbose: print "no bins entered!"
datmean = dat.mean()
datstd = stdr(dat, nsigma=3)
nunique = len(np.unique(dat.ravel()))
#pdb.set_trace()
if nunique > len(dat)/20.:
dobin = False
elif nunique>1:
dobin = True
bins = linspace(dat.min(), dat.max(), nunique/2)
else:
dobin = True
bins = np.array([dat[0]-1, dat[0]+1])
if plotalot or verbose:
print "dat.mean, dat.std>>" + str((dat.mean(), dat.std()))
#if plotalot:
# figure(); hout = hist(dat[datIndex],bins)
#else:
if dobin:
binwidth = mean(bins[1::]-bins[:-1])
bincenter = 0.5*(bins[1::]+bins[:-1])
datIndex = (dat>=bins.min()) * (dat<=bins.max())
hout = histogram(dat[datIndex], bins) #,new=True)
gy = hout[0]
erry = sqrt(gy)
usableIndex = gy>0
if usableIndex.size==0:
return badval, badval
eff_binwidth = mean(bins[usableIndex][1::]-bins[usableIndex][:-1])
guess = [gy.sum()*eff_binwidth, std(dat[datIndex]), median(dat[datIndex])]
if 1.0*usableIndex.sum()/usableIndex.size < 0.5:
out = guess
else:
out = optimize.fmin(gaussianChiSquared, guess, \
args=(bincenter[usableIndex],gy[usableIndex], erry[usableIndex]), \
disp=plotalot)
if plotalot:
from pylab import figure, errorbar, plot, colorbar, title
from nsdata import imshow
print 'guess>>',guess
print 'fit>>',out
figure()
imshow(im); colorbar()
figure()
errorbar(bincenter[usableIndex], gy[usableIndex], erry[usableIndex], fmt='ob')
plot(bincenter, gaussian(out, bincenter),'-r', linewidth=2)
title('Mean: %f, Std. Dev.: %f' % (out[2], out[1]))
ret = out[2], out[1]/sqrt(ndat)
else:
ret = datmean, datstd/sqrt(ndat)
return ret
def estbg(im, mask=None, bins=None, plotalot=False, rout=(3,200), badval=nan, verbose=False):
"""Estimate the background value of a masked image via histogram fitting.
INPUTS:
im -- numpy array. Input image.
OPTIONAL INPUTS:
mask -- numpy array. logical mask, False/0 in regions to ignore
bins -- sequence. edges of bins to pass to HIST
plotalot -- bool. Plot the histogram and fit.
rout -- 2-tuple of (nsigma, niter) for analysis.removeoutliers.
Set to (Inf, 0) to not cut any outliers.
badval -- value returned when things go wrong.
OUTPUT:
b, s_b -- tuple of (background, error on background) from gaussian fit.
Note that the error is analagous to the standard deviation on the mean
COMMENTS:
The fit parameters appear to be robust across a fairly wide range of bin sizes. """
# 2009-09-02 17:13 IJC: Created!
# 2009-09-04 15:07 IJC: Added RemoveOutliers option. Use only non-empty bins in fit.
# 2009-09-08 15:32 IJC: Error returned is now divided by sqrt(N) for SDOM
# 2009-11-03 00:16 IJC: Improved guess for gaussian dispersion
# 2011-05-18 11:47 IJMC: Moved (e)gaussian imports to analysis.
# 2012-01-01 21:04 IJMC: Added badval option
# 2012-08-15 17:45 IJMC: Numpy's new histogram no longer accepts 'new' keyword
# 2013-03-20 08:22 IJMC: Now works better even for small numbers
# of pixels; thanks to A. Weigel @
# ETH-Zurich for catching this!
# 2014-08-29 10:05 IJMC: Added verbosity flag.
# 2014-09-25 14:29 IJMC: Added check for no usability
from numpy import histogram, mean, median, sqrt, linspace, isfinite, ones,std
from pylab import find
from scipy import optimize
from analysis import removeoutliers, egaussian, gaussian, stdr
if plotalot:
from pylab import figure, errorbar, plot, colorbar, title, hist, mean, std
#from analysis import imshow
def gaussianChiSquared(guess, x, y, err):
return (egaussian(guess, x, y, e=err)**2).sum()
if mask==None:
mask = ones(im.shape)
dat = im.ravel()[find(mask<>0)]
if plotalot:
figure(); plot(im.ravel()); plot(dat)
print mean(dat), std(dat), rout[0]*std(dat)
print len(dat), (abs(dat-mean(dat))<(rout[0]*std(dat))).sum()
figure(); plot(dat-mean(dat));
plot([0,len(dat)], [rout[0]*std(dat),rout[0]*std(dat)],'--k')
plot([0,len(dat)], [-rout[0]*std(dat),-rout[0]*std(dat)],'--k')
dat = removeoutliers(dat, rout[0], remove='both', center='mean', niter=rout[1], verbose=plotalot)
ndat = len(dat)
if ndat==0:
if verbose>0: print "No data to work with!"
return (badval, badval)
if bins==None:
if plotalot or verbose: print "no bins entered!"
datmean = dat.mean()
datstd = stdr(dat, nsigma=3)
nunique = len(np.unique(dat.ravel()))
#pdb.set_trace()
if nunique > len(dat)/20.:
dobin = False
elif nunique>1:
dobin = True
bins = linspace(dat.min(), dat.max(), nunique/2)
else:
dobin = True
bins = np.array([dat[0]-1, dat[0]+1])
if plotalot or verbose:
print "dat.mean, dat.std>>" + str((dat.mean(), dat.std()))
#if plotalot:
# figure(); hout = hist(dat[datIndex],bins)
#else:
if dobin:
binwidth = mean(bins[1::]-bins[:-1])
bincenter = 0.5*(bins[1::]+bins[:-1])
datIndex = (dat>=bins.min()) * (dat<=bins.max())
hout = histogram(dat[datIndex], bins) #,new=True)
gy = hout[0]
erry = sqrt(gy)
usableIndex = gy>0
if usableIndex.size==0:
return badval, badval
eff_binwidth = mean(bins[usableIndex][1::]-bins[usableIndex][:-1])
guess = [gy.sum()*eff_binwidth, std(dat[datIndex]), median(dat[datIndex])]
if 1.0*usableIndex.sum()/usableIndex.size < 0.5:
out = guess
else:
out = optimize.fmin(gaussianChiSquared, guess, \
args=(bincenter[usableIndex],gy[usableIndex], erry[usableIndex]), \
disp=plotalot)
if plotalot:
from pylab import figure, errorbar, plot, colorbar, title
from nsdata import imshow
print 'guess>>',guess
print 'fit>>',out
figure()
imshow(im); colorbar()
figure()
errorbar(bincenter[usableIndex], gy[usableIndex], erry[usableIndex], fmt='ob')
plot(bincenter, gaussian(out, bincenter),'-r', linewidth=2)
title('Mean: %f, Std. Dev.: %f' % (out[2], out[1]))
ret = out[2], out[1]/sqrt(ndat)
else:
ret = datmean, datstd/sqrt(ndat)
return ret
