Skip to main content
. Author manuscript; available in PMC: 2015 Feb 11.
Published in final edited form as: Neural Comput. 2015 Jan;27(1):151–170. doi: 10.1162/NECO_a_00690

Figure 3.

Figure 3

Illustrative fMRI dataset; diagnostics for eliciting the regularity parameter λ based on technical replicate data and retrospective inspection of the posterior. (a) Here two subjects each provided four technical replicate datasets. The DAGs shown are joint MAP estimates for varying λ, such that replicates were assumed to be exchangeable but subjects were treated independently. As λ is increased the DAGs corresponding to technical replicates become increasingly similar until they are identical at λ ≥ 17. (b) Here we plot the total SHD between DAGs corresponding to technical replicates against the regularity parameter λ. The dashed line indicates the value λ = 4 that reduces the between-subject variability by approximately 50%. (c) Comparing the Bayes factor corresponding to model λ = 4 against independent estimation (λ = 0) and estimation that forces all DAGs to be identical (λ ≥ 17).