def linefit(x,y,weights=None):
"""
Parameters
----------
y: 1D numpy array
The data to be fitted
x: 1D numpy array
The x values of the y array. x and y must
have the same shape.
weights: 1D numpy array, must have the same shape as x and y
weight values
Examples
--------
>>> x=N.array([-5, -4 ,-3 ,-2 ,-1, 0, 1, 2, 3, 4, 5])
>>> y=N.array([1, 5, 4, 7, 10, 8, 9, 13, 14, 13, 18])
>>> around(linefit(x,y), decimals=5)
array([ 9.27273, 1.43636])
>>> x=N.array([1.3,1.3,2.0,2.0,2.7,3.3,3.3,3.7,3.7,4.,4.,4.,4.7,4.7,5.,5.3,5.3,5.3,5.7,6.,6.,6.3,6.7])
>>> y = N.array([2.3,1.8,2.8,1.5,2.2,3.8,1.8,3.7,1.7,2.8,2.8,2.2,3.2,1.9,1.8,3.5,2.8,2.1,3.4,3.2,3.,3.,5.9])
>>> around(linefit(x,y), decimals=5)
array([ 1.42564, 0.31579])
"""
if numerixenv.check_input(x) or numerixenv.check_input(y):
raise ValueError, "Input is a NumArray array. This version of %s requires a Numpy array\n" % __name__
if len(x) != len(y):
print "Error: X and Y must have equal size\n"
return
n = len(x)
w = N.zeros((n,n)).astype(N.float)
if weights == None:
for i in N.arange(n):
w[i,i] = 1
else:
if len(weights) != n:
print "Error: Weights must have the same size as X and Y.\n"
return
for i in N.arange(n):
w[i,i] = weights[i]
x = x.astype(N.float)
y = y.astype(N.float)
# take the weighted avg for calculatiing the covarince
Xavg = N.sum(N.dot(w,x)) / N.sum(w.diagonal())
Yavg = N.sum(N.dot(w,y)) / N.sum(w.diagonal())
xm = x - Xavg
xmt = N.transpose(xm)
ym = y - Yavg
b1 = N.dot(xmt,N.dot(w,ym)) / N.dot(xmt ,N.dot(w,xm))
b0 = Yavg - b1 * Xavg
return b0, b1
def gfit1d(y, x=None, err = None, weights=None, par=None, parinfo=None,
maxiter=200, quiet=0):
"""
Return the gaussian fit as an object.
Parameters
----------
y: 1D Numarray array
The data to be fitted
x: 1D Numarray array
(optional) The x values of the y array. x and y must
have the same shape.
err: 1D Numarray array
(optional) 1D array with measurement errors, must be
the same shape as y
weights: 1D Numarray array
(optiional) 1D array with weights, must be the same
shape as y
par: List
(optional) Starting values for the parameters to be fitted
parinfo: Dictionary of lists
(optional) provides additional information for the
parameters. For a detailed description see nmpfit.py.
Parinfo can be used to limit parameters or keep
some of them fixed.
maxiter: number
Maximum number of iterations to perform
Default: 200
quiet: number
if set to 1, nmpfit does not print to the screen
Default: 0
Examples
--------
>>> x=N.arange(10,20, 0.1)
>>> y= 10*N.e**(-(x-15)**2/4)
>>> print gfit1d(y,x=x, maxiter=20,quiet=1).params
[ 10. 15. 1.41421356]
"""
if numerixenv.check_input(x) or numerixenv.check_input(y):
raise ValueError, "Input is a NumArray array. This version of %s requires a Numpy array\n" % __name__
y = y.astype(N.float)
if weights != None:
weights = weights.astype(N.float)
if err != None:
err = err.astype(N.float)
if x == None and len(y.shape)==1 :
x = N.arange(len(y)).astype(N.float)
if x.shape != y.shape:
print "input arrays X and Y must be of equal shape.\n"
return
fa = {'x':x, 'y':y, 'err':err, 'weights':weights}
if par != None:
p = par
else:
ysigma = y.std()
ind = N.nonzero(y > ysigma)[0]
if len(ind) != 0:
xind = int(ind.mean())
p2 = x[xind]
p1 = y[xind]
p3 = 1.0
else:
ymax = y.max()
ymin = y.min()
ymean= y.mean()
if (ymax - ymean) > (abs(ymin - ymean)):
p1 = ymax
else: p1 = ymin
ind = (N.nonzero(y == p1))[0]
p2 = x.mean()
p3 = 1.
p = [p1, p2, p3]
m=nmpfit.mpfit(_gauss_funct, p,parinfo = parinfo, functkw=fa,
maxiter=maxiter, quiet=quiet)
if (m.status <=0): print 'error message = ', m.errmsg
return m
