Fixed: the calculation of the data frame containing the indices for the nested sums was wrong for dimension p>3.

Correction in 'Arguments' section: 'eps' is the precision of the partial derivative (not of the KL divergence)