The condition of the latest versions of the ELMy H-mode and L-mode databases have been re-examined in view of their sensitivity to errors in the absorbed heating power and stored energy. It is shown that there is bias in the OLS regression for some of the variables. These short comings are overcome by the use of an error in variables technique and new scalings are derived. These give a very similar performance to existing scalings for ITER at the standard bn of 1.8, but much improved performance at higher bn.