We want to be able to compare the resulting inference of these different approaches. It is difficult to compare different t-statistics with different DOF. Therefore, for each of the different inference approaches we convert to the probability of the contrast being greater than zero. This provides us with a measure which we can compare directly between the different approaches. We represent this probability as a z-statistic by ensuring that the area under one tail of a standardised (zero mean and standard deviation of one) Normal distribution corresponds to that probability. In section 6.3 we will explore the possibility of using these z-statistics to mimic null-hypothesis frequentist inference.