Centroid calculation is incorrect.

[Andrew Nelson]

import numpy as np
from peakutils import centroid

y = np.ones(10.)
x = np.arange(10.)
c = centroid(x, y)
assert(c == 4.5)

y = np.ones(3.)
x = np.array([0., 1., 9.])
c = centroid(x, y)
assert(c == 4.5)

A better calculation is:

def centroid(y, x=None, dx=1.):
    '''Computes the centroid for the specified data.

    Parameters
    ----------
    y : array_like
        Array whose centroid is to be calculated.
    x : array_like, optional
        The points at which y is sampled.
    Returns
    -------
    (centroid, sd)
        Centroid and standard deviation of the data.
    '''
    yt = np.array(y)

    if x is None:
        x = np.arange(yt.size, dtype='float') * dx

    normaliser = simps(yt, x)
    centroid = simps(x * yt, x) / normaliser
    var = simps((x - centroid)**2 * yt, x) / normaliser
    return centroid, np.sqrt(var)

[Lucas Hermann Negri]

Thanks. I've added your implementation as centroid2. The original one is not incorrect, but its formulation indeed does not work for your test data.