Direct Age Standardization fails for empty regions

[ljwolf]

I think there's a minor bug in pysal.esda.smoothing.direct_age_standardization:

for i in range(len(adjusted_r)):
    if adjusted_r[i] == 0:
        upper = .5 * scipy.stats.chi2(1-0.5 * alpha)
        lower = 0
    else:
        lower = scipy.stats.gamma.ppf(.05 * alpha, g_a[i],. scale=g_b[i])
        upper = scipy.stats.gamma.ppf(1 - .05*alpha, g_a_[k], scale=g_b_k[i]

I think that should be scipy.stats.chi2.ppf, as is the case for the gamma function. As it stands, when adjusted_r[i] == 0, we get a TypeError, since calling chi2(parameters) directly creates a frozen random variable, which can't be multiplied.

The traceback that occurs is:

/home/ljw/dev/pysal/pysal/esda/smoothing.pyc in __init__(self, e, b, w, s, alpha)
    978         e_n = np.array(e_n).reshape((1, t), order='F')[0]
    979         b_n = np.array(b_n).reshape((1, t), order='F')[0]
--> 980         r = direct_age_standardization(e_n, b_n, s, w.n, alpha=alpha)
    981         self.r = np.array([i[0] for i in r])
    982         w.transform = 'o'

/home/ljw/dev/pysal/pysal/esda/smoothing.pyc in direct_age_standardization(e, b, s, n, alpha)
    283     for i in range(len(adjusted_r)):
    284         if adjusted_r[i] == 0:
--> 285             upper = 0.5 * chi2(1 - 0.5 * alpha)
    286             lower = 0.0
    287         else:

TypeError: unsupported operand type(s) for *: 'float' and 'rv_frozen'

Someone with a knowledge of smoothing should probably review.

[ljwolf]

should be resolved in #827