1

I am working on my first assignment using bnlearn package to perform EDA. I have created a network using hill climb (hc) in R with all the default values.

BUT there are few nodes in Bayesian Network which does NOT have any predecessor or successor node in the Directed acyclic graph(DAG) created.

What is the way to check the score or criteria for NODE not having any connected node. How can we explain this.

Thanks a lot

Maddy
  • 23
  • 6
  • The total score of the network sums up the local scores at each node. If your net is discrete, you can calculate the log-likelihood of a (local) single node, even if it has no edges, by using the [multinomial pdf](https://en.wikipedia.org/wiki/Multinomial_distribution). Try and calculate it for the net: `g = hc(learning.test[ "F" ])` , your answer should give `logLik(g, learning.test[ "F" ])` (ps it may be off by a constant) . Also, if a node has no edges it just means that it is independent from the others - it happens. – user20650 May 24 '18 at 19:42

0 Answers0