I have been doing data driven
causal inference from Bayesian Networks in R using
bnlearn package on a random categorical dataset.
The data is used to learn the structure and parameters of model.
Below is the code for learning structure of Bayesian network.
boot_hc <- boot.strength(data = df, R = 100, m = nrow(df), algorithm = "hc", algorithm.args = list(score="bic"), cpdag = F, debug = T)
bn_hc <- averaged.network(strength = boot_hc, threshold = 0.85)
The BIC score for the above DAG is
-339005.6 computed using
> score(x = bn, data = df, type = "bic")
My question is
How can I normalize the BIC score (or any other score function) in order to get more understanding of quality of fit of learned structure ?