geary optimization

[ljwolf]

we should be able to exploit numpy.subtract.outer and elementwise matrix multiplication to do the Geary statistic in a much faster fashion than our current list comprehension. Something like:

numerator = (numpy.subtract.outer(y,y)**2).multiply(w.sparse).sum()

[ljwolf]

or, even better, use sklearn.metrics.pairwise and we might be able to get something like to @lanselin's multivariate geary with minimal change (though still might need to extend the permutation inference stuff for it.)

[sjsrey]

we should be able to exploit numpy.subtract.outer and elementwise matrix multiplication to do the Geary statistic in a much faster fashion than our current list comprehension. Something like:

numerator = (numpy.subtract.outer(y,y)**2).multiply(w.sparse).sum()

the outer will generate a dense nxn matrix though?

[jGaboardi]

Closed by #114.