Fig. 1.

Percentage of non-converged models from selected methods. Methods in which less than 1% of models failed to converge are not included in the figure. For NBMM-LP, which uses a likelihood ratio test, the solid portion of the bar represents the proportion of models in which the full model did not converge and the transparent portion represents genes for which the reduced model for one or more tests failed to converge in which case results for those tests could not be obtained