output posterior mean-squared value given prior mixture m and data
comp_postmean2(m, data)
m
mixture distribution with k components
data
details depend on the model