R Statistics Cookbook

Conditional independence tests

In the previous recipe, we first started with an expert-defined model, and we then showed how to do it using an automatic approach. But this last one didn't work too well for our case, as the resulting model didn't make much sense. We finally decided to include some arcs, which were then complemented by the automatic algorithm using the whitelist= parameter.

Independence tests are used by the automatic algorithms to find out reasonable structures, but they can also be employed by the user in the expert decision process. More concretely, we might want to add a new node, and using these independence tests, we can decide whether adding it makes sense or not.