Phase space exploration

This is a tool we have not used much, because it is more a way of checking that everything is ok. Nevertheless, it can be used to look at the information the data gives us form a different angle. Basically, it consists in taking each variable,  giving it a particular value, and looking at how all the other variables behave.
As you can imagine, this would generate an incredibly high number of pictures that could not possibly be looked at in a way that could prove useful. That is why we usually suggest to look only at a subset of the variables while working with the bayesian network. But if we make the compromise of taking the averages of the various variables, we can study how each variable behaves - on average - when we change other variables, according to our model. This is very interesting because it gives us a quantitative information again, but it has the disadvantage of only showing averages. Averages are good to have an overall idea of the situation, but if we want to have a better idea of what is happening, we should be using distributions, which is how the bayesian network works by default.