Fig. 4.

For a given metabolite peak, the measured response can be plotted against injection order (excluding conditioning samples and blanks) and the time varying systematic variation in metabolite response observed (a). The systematic variation can be modelled, in this case using a regularised cubic spline with a smoothing parameter. The optimal smoothing parameter value is the one with the lowest cross-validated error (b). The ‘correction curve’ can then be subtracted from the raw data (c). Accurate measures of precision after the correction can then be calculated (d). Red squares are QC samples, blue circles are study samples