Testing the log fold change with regard to a threshold versus categorisation into relevance classes. Comparison of testing the log fold change with regard to a threshold (McCarthy and Smyth ) versus categorisation of confidence intervals into relevance classes by our approach. Both methods show a strong correlation. In particular, nearly all significant p-values fall in the strongest relevance category D. The comparison was performed using the data of Lips et al.  and a relevance threshold of ρ = 0.5.