Improved classification accuracy in 1- and 2-dimensional NMR metabolomics data using the variance stabilising generalised logari