Better estimates of the distribution’s center arise from averaging over more data.

$$Q(\lambda)= \frac{1}{M} \sum_{j=1}^{M}O_j$$

This is how we can improve the vector-valued posterior density.