sN - sTmix Model Sample Clauses

sN - sTmix Model. The distribution of z values can sometimes exhibit skewness. As mentioned in Section 3.2.5, this asymmetry is mostly due to errors that are asymmetric and/or discontin- uously distributed. For example, Xxxxx-Xxxx´andez et al., (2007)[98] found that the distribution of 18O labeled log-2 ratios was not correctly centered on zero but was sig- nificantly biased toward the non-labeled sample, displaying a significant asymmetry comprising of an extended right tail. The authors found that this asymmetry was due to the presence of a small number of peptides having a low but significant proportion of non-labeled species. These types of distortions in the symmetry of the distribution is often not effectively captured by mixture models that use symmetric components, unless allowance is made for the asymmetry introduced by the skewness. In the sN- sTmix model setup, we extend the Nmix-Tmix model by allowing the component normal and t distributions to have an additional shape parameter that determines skewness. This provides a more flexible approach to the fitting of asymmetric sub- classes that exhibit varying degrees of skewness. For the most part, the additional flexibility provided by the skew-Normal distribution allows us to fit f0(z) accurately with only a single component model. Similarly, the additional flexibility afforded by the skew-t distribution usually results in fewer mixture components needed in the fitting of f (z).