Comparing both methods showed the resulting posterior distributions for the GP to be near identical,
with the posterior means m on average within 0.3% (relative error) of each other, and the credible bands
within 1.2% of each other. This indicates stability of the GP estimates given slightly different hyperparameters.