Would it be possible to add other options for the collocation statistic? I am thinking particularly of log likelihood.