MCMC error estimates for a two-component Gaussian fit. The synthetic spectrum consists of two Gaussian components and white noise with σ_T = 10% relative to the peak value in the spectrum. The corner plot shows the parameter correlations, where the colours of the 2D histogram correspond to the logarithm of probability (MCMC samples per pixel), as indicated by the colour bar on top. The red crosses correspond to the least-squares solution. The diagonal frames show the histograms for the individual parameters, and the dashed black and red lines correspond to the true values (before observational noise is added) and to the least squares fit, respectively. The frame on the right shows this particular fitted spectrum (black line) and the two components corresponding to the χ² minimum.