Sometimes when the parameters of a model are entered manually by experts instead of learnt from data, two or more experts may disagree on the values to be entered

In scenarios such as this, a new discrete node can be added to the network, each state of which represents an expert, as shown below:

Expert Disagreement Network

The node that the experts disagree on could be discrete or continuous, but the node representing the experts is always discrete.

The distribution required for the Expert node, allows us to give a weight to each expert.

They could be given equal weight, or some experts might be considered more important, as in the distribution shown below.

Expert Disagreement Expert Distribution

The distribution required for the Market node in this example, allows values to specified for each expert as shown below:

Expert Disagreement Market Distribution