Figure 2: Distribution of ln(eGFR) values across all measurements for all patients within our dataset who have been diagnosed with CKD between stages 3 to 5. The solid line is the best-fitting normal distribution. If the dependent variable (i.e. eGFR) is subtracted from a constant value (i.e. the highest eGFR value found in our dataset), then it can be assumed that this transformed dependent variable (107-eGFR) will follow a standard gamma distribution. As can be seen from figure 3, even the distribution is still not normal, distribution in figure 3 is closer to normal compared to figure 2 and, hence, the assumption of gamma distribution when using the transformed eGFR value rather than the eGFR value itself is better, which is used in GLMM model 5.