bayesdistance.py 文件源码

def log_bf(p, s):
    """
    log10 of the multi-way Bayes factor, see eq.(18)

    p: separations matrix (NxN matrix of arrays)
    s: errors (list of N arrays)
    """
    n = len(s)
    # precision parameter w = 1/sigma^2
    w = [numpy.asarray(si, dtype=numpy.float)**-2. for si in s]
    norm = (n - 1) * log(2) + 2 * (n - 1) * log_arcsec2rad

    wsum = numpy.sum(w, axis=0)
    s = numpy.sum(log(w), axis=0) - log(wsum)
    q = 0
    for i, wi in enumerate(w):
        for j, wj in enumerate(w):
            if i < j:
                q += wi * wj * p[i][j]**2
    exponent = - q / 2 / wsum
    return (norm + s + exponent) * log10(e)