5.3 Markov chain Monte Carlo inversions – median fits and
convergence diagnostics
The model simulations used for illustration in section 5.2 were
conducted with unconstrained values of \(\xi\). To obtain constraints on \(\xi\) and, importantly, how it varies as a function of \(\phi\), MCMC
inversions were performed on the experiments of Hoyos et al.(2022) and the olivine centrifuge experiments using eq. (16). For each
inversion, three MCMC realizations (chains) were performed. Each of the
three realizations per group were initiated with different random values
of \(m^{\text{ini}}\). Examples of the fits calculated using the most
likely combination of parameters (the vector \(m^{\text{med}}\) of
median values for each parameter) obtained by the inversions are shown
in Fig. 5 and 6 for the low T + P experiments of Hoyos et al. (2022).