9.3 Multiple Coins from Multiple Mints 219 proportion-correct actually yielded by the practitioners, which were 0.1 and 0.8, respectively. Although the shrinkage pulls the individual estimates toward the group average, the shrinkage also increases the certainty of the estimate of the group average. In other words, when is larger, the distribution of µ gets narrower. This effect on the spread of µ can be seen in the upper-left panel of Figure 9.13, which shows a scatter plot of µ and for a similar toy design. Thus, the consistency across subjects, which shrinks individual estimates, also enhances the specificity of the group estimate. 9.2.5 Number of Coins and Flips per Coin When we collect more data, our estimate of the model parameters becomes more certain. For example, in the previous section's investigation of thera- peutic touch (TT), we know that the sensitivity of the experiment could have been larger if more data were collected. Indeed, if enough data were collected, maybe we could conclude that TT practitioners can, on average, sense the experimenter's hand, albeit only weakly. The new twist to this old theme is that here we have two ways to get more