Fig. 2: Robust linear regression effectively estimates spillover coefficients.

Each row a–d, e–h, i–l, m–p shows a compensation example from the MM1 dataset, with the primary and secondary channels, indicated, respectively, in the y-axes and x-axes. Compensation results are displayed using positive and negative populations (first column (a, e, i, m), bi-exponential scale; second column (b, f, j, n), linear scale), and robust linear regression (third column (c, g, k, o), bi-exponential scale; fourth column (d, h, l, p), linear scale). The linear relationship between the levels of fluorescence is not visible in bi-exponential scale, but it is very clear in linear scale. Uncompensated data is displayed in blue and compensated data in black. Dim points correspond to gated-out events, not used in the calculation. Lines in the second and fourth columns (linear scale) show regressions of uncompensated (blue) and compensated (black) data. The slope coefficient of the latter provided the compensation error (number at the bottom right of each panel). Vertical green dashed lines are shown as a reference for perfectly compensated data. Raw dataset is available at FlowRepository with ID FR-FCM-Z2SS (MM1) [https://flowrepository.org/id/FR-FCM-Z2SS].