empiricaldistribution.py 文件源码

def __init__(self, data, names=None):

        if names is None:
            self.names = range(data.shape[1])
        else:
            assert (len(names) == self.NVAR), 'Passed-in names length must equal number of data columns'
            self.names = names

        self.NROW = data.shape[0]
        self.NVAR = data.shape[1]
        self.bins = [len(np.unique(data[:,n])) for n in range(self.NVAR)]


        hist,_ = np.histogramdd(data, bins=self.bins)
        self.counts = hist
        self.joint = (hist / hist.sum()) + 1e-3

        ## COMPUTE MARGINAL FOR EACH VARIABLE ##
        #_range = range(self.NVAR)
        #for i,rv in enumerate(self.names):
        #   _axis = copy(_range)
        #   _axis.remove(i)
        #   self.marginal[rv] = np.sum(self.joint,axis=_axis)

        #self.marginal = dict([(rv, np.sum(self.joint,axis=i)) for i,rv in enumerate(self.names)])

        self.cache = {}