[SciPy-User] How to fit data with errorbars

josef.pktd@gmai... josef.pktd@gmai...
Wed Feb 17 11:30:33 CST 2010


On Wed, Feb 17, 2010 at 12:12 PM, Joe Harrington <jh@physics.ucf.edu> wrote:
>> I have some data with some errobars that I need to fit to a line.
>
> Actually, you should fit the line; please don't change the data! ;-)
>
> The routine below does the job.  I was surprised not to find this
> routine in numpy or scipy as it's very standard (I'm still convinced I
> must have missed it, so somebody please embarrass me and point it
> out).

scipy.stats.linregress for linear fit in one explanatory variable, but
ols only, there are no weights.
optimize.curve_curvefit uses weights but is intended for non-linear
(in parameters), and scipy.odr is very flexible.

And of course there is scikits.statsmodels for this and much more.

Josef

> The routine has a number of parameters that are standard in other
> numerical packages, such as IDL, for *every* fitting routine.  We
> should consider adding these parameters to all our fitting routines,
> as well, since having them encourages people to fit (more) properly
> (i.e., actually assess the goodness of fit, etc.).
>
> --jh--
>
> import numpy as np
> import scipy.special as ss
>
> def linfit(y, x=None, y_unc=None):
>  '''
>    Fits a line to 2D data, optionally with errors in y.
>
>    The method is robust to roundoff error.
>
>    Parameters
>    ----------
>    y: ndarray
>        Ordinates, any shape.
>    x: ndarray
>        Abcissas, same shape.  Defaults to np.indices(y.length))
>    y_unc: ndarray
>        Uncertainties in y.  If scalar or 1-element array, applied
>        uniformly to all y values.  [NOT IMPLEMENTED YET!]  Must be
>        positive.
>
>    Returns
>    -------
>    a: scalar
>        0 Fitted intercept
>    b: scalar
>        1 Fitted slope
>    a_unc: scalar
>        2 Uncertainty of fitted intercept
>    b_unc: scalar
>        3 Uncertainty of fitted slope
>    chisq: scalar
>        4 Chi-squared
>    prob: scalar
>        5 Probability of finding worse Chi-squared for this model with
>          these uncertainties.
>    covar: ndarray
>        6 Covariance matrix: [[a_unc**2,  covar_ab],
>                              [covar_ab,  b_unc**2]]
>    yfit: ndarray
>        7 Model array calculated for our abcissas
>
>    Notes
>    -----
>
>    If prob > 0.1, you can believe the fit.  If prob > 0.001 and the
>    errors are not Gaussian, you could believe the fit.  Otherwise
>    do not believe it.
>
>    See Also
>    --------
>    Press, et al., Numerical Recipes in C, 2nd ed, section 15.2,
>    or any standard data analysis text.
>
>    Examples
>    --------
>    >>> import linfit
>
>    >>> a = 1.
>    >>> b = 2.
>    >>> nx = 10
>    >>> x = np.arange(10, dtype='float')
>    >>> y = a + b * x
>    >>> y_unc = numpy.ones(nx)
>    >>> y[::2]  += 1
>    >>> y[1::2] -= 1
>    >>> a, b, sa, sb, chisq, prob, covar, yfit = linfit.linfit(y, x, y_unc)
>    >>> print(a, b, sa, sb, chisq, prob, covar, yfit)
> (1.272727272727272, 1.9393939393939394, 0.58775381364525869, 0.11009637651263605, 9.6969696969696937, 0.28694204178663996, array([[ 0.34545455, -0.05454545],
>       [-0.05454545,  0.01212121]]), array([  1.27272727,   3.21212121,   5.15151515,   7.09090909,
>         9.03030303,  10.96969697,  12.90909091,  14.84848485,
>        16.78787879,  18.72727273]))
>
>    Revisons
>    --------
>    2007-09-23 0.1  jh@physics.ucf.edu  Initial version
>    2007-09-25 0.2  jh@physics.ucf.edu  Fixed bug reported by Kevin Stevenson.
>    2008-10-09 0.3  jh@physics.ucf.edu  Fixed doc bug.
>    2009-10-01 0.4  jh@physics.ucf.edu  Updated docstring, imports.
>  '''
>  # standardize and test inputs
>  if x == None:
>    x = np.indices(y.length, dtype=y.dtype)
>    x.shape = y.shape
>
>  if y_unc == None:
>      y_unc = np.ones(y.shape, dtype=y.dtype)
>
>  # NR Eq. 15.2.4
>  ryu2  = 1. / y_unc**2
>  S     = np.sum(1.    * ryu2)
>  Sx    = np.sum(x     * ryu2)
>  Sy    = np.sum(y     * ryu2)
>  # Sxx = np.sum(x**2  * ryu2) # not used in the robust method
>  # Sxy = np.sum(x * y * ryu2) # not used in the robust method
>
>  # NR Eq. 15.2.15 - 15.2.18 (i.e., the robust method)
>  t = 1. / y_unc * (x - Sx / S)
>  Stt = np.sum(t**2)
>
>  b = 1. / Stt * np.sum(t * y / y_unc)
>  a = (Sy - Sx * b) / S
>
>  covab = -Sx / (S * Stt)                  # NR Eq. 15.2.21
>
>  sa = np.sqrt(1. / S * (1. - Sx * covab)) # NR Eq. 15.2.19
>  sb = np.sqrt(1. / Stt)                   # NR Eq. 15.2.20
>
>  rab = covab / (sa * sb)                  # NR Eq. 15.2.22
>
>  covar = np.array([[sa**2, covab],
>                    [covab, sb**2]])
>
>  yfit = a + b * x
>  chisq = np.sum( ((y - yfit) / y_unc)**2 )
>
>  prob = 1. - ss.gammainc( (y.size - 2.) / 2., chisq / 2.)
>
>  return a, b, sa, sb, chisq, prob, covar, yfit
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>


More information about the SciPy-User mailing list