The Gaussian curves of figure 1 become the quadratics of figure 2, with the most likely
values for x being at the minimum of the total curve. So the Bayesian combination of
Gaussian pdfs gives the same "best" analysis as a weighted least-squares best fit to the data.