calculates parameter-optimized fit (including LL) using a dirichlet multinomial model given a matrix of data (k), a model matrix with the features (mm), and a concentration parameter.
calculates a parameter-non-optimized log-likelihood using a dirichlet multinomial model given a matrix of data (k), a model matrix with the features (mm), and a concentration parameter.
This is useful for data sets that fail to converge the optimization process. e.g. those with not so many samples.