Univariate statistical analysis of gas chromatography – mass spectrometry fingerprints analyses

Abstract Gas Chromatography - Mass Spectrometry (GC-MS) has been used for a long time in fingerprint analysis. We present a workflow of univariate statistical treatment of compound by considering their type of response variables. Two data sources were used: (i) comparative data from two Brazilian Amazon soils, and (ii) the Nitrogen-dose response experiment involving two Ilex paraguariensis clones. During type of response variables selection, the following assumptions were tested: normality and homogeneity of variances. After defining a strategy to select the type of response variables, the compounds were classified according to the statistical test that must be used to evaluate them: analysis of variance (ANOVA, LM), generalized linear model (GLM), and a non-parametric (NP) test. The developed workflow allows individual compound and class comparisons, and a couple examples that illustrate a wider range of similar datasets are open to the readers to test either their own data or ours.