next up previous
Next: Conclusions and Discussion Up: Null Data - Pseudo Previous: Methods

Results

The results are shown as a p-p plot in figure 14. There is good correspondence between the nominal frequentist rate and the rate obtained experimentally using the fully Bayesian spatio-temporal inference. This is because we have a non-informative prior on the activation height parameter, $ a_i$.
Figure 14: (a) Probability-probability (p-p) plot and (b) Log probability-log probability plot, comparing nominal frequentist ``pseudo-FPR'', $ H$, against the experimentally obtained proportion of voxels with $ p(a_i>0\vert y)>1-H$, obtained from a null/rest dataset for the random ISI single-event design using the full model.
\begin{figure}
\centering
\begin{tabular}{cc}
\psfig{file=fig14a.ps,width=0....
...=fig14b.ps,width=0.22\textwidth}\\
(a)&(b)\\
\end{tabular}
\end{figure}