A Comparative Analysis of Four PLS-DA Diagnostic Statistics in the Application of Metabolomics
Objective To compare the statistical power and stability of four PLS-DA diagnostic statistics in the analysis of metabolomic data.Methods The simulated data and realistic data were analyzed based on the PLS-DA validation strategy of double cross validation and permutation test in conjunction with four diagnostic statistics.Results AUC showed higher statistical power than misclassification number(rate),Q2 and DQ2 ;in the meanwhile,AUC was more stable than the other diagnostic statistics.Conclusion AUC is a stable and effective diagnostic statistic in the validation of PLS-DA models,and is recommended as the preferred diagnostic statistic in the PLS-DA analysis of metabolomic studies.
PLS-DAMetabolomicsDiagnostic statisticDouble cross validationPermutation test
NETL
NSTL
万方数据
维普
