Automatic calculation of expected value of decision node

The front end.
Post Reply
regnier
Posts: 4
Joined: Fri Jul 31, 2009 8:21 pm

Automatic calculation of expected value of decision node

Post by regnier »

Does GeNIe have a way to automatically calculate the expected value of the first decision node in an influence diagram, if it is the child of an uncertainty node, or do I have to take the optimal expected values for each outcome of the uncertainty node and take their expectation by multiplying by the probabilities of each outcome, externally to GeNIe?

Thanks!
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Re: Automatic calculation of expected value of decision node

Post by marek [BayesFusion] »

regnier wrote:Does GeNIe have a way to automatically calculate the expected value of the first decision node in an influence diagram, if it is the child of an uncertainty node, or do I have to take the optimal expected values for each outcome of the uncertainty node and take their expectation by multiplying by the probabilities of each outcome, externally to GeNIe?
Sorry for the delay in answering. You can easily get what you want by removing the arc from the chance node to the decision node. You can do it temporarily (i.e., restore it afterwards), as there are no consequences for the definition of the decision node and nothing will change there.

Please keep in mind that an arc from a chance node into a decision node means that the outcome of the chance node will be observed *before* the decision is made, so this is why you have to do the trick with temorary removal, if you really want it both, have the cake and eat it :-). I hope this helps.
Cheers,

Marek
regnier
Posts: 4
Joined: Fri Jul 31, 2009 8:21 pm

Post by regnier »

This doesn't actually give me what I want. What I would like is the expected value of the optimal decision at the decision node, when that decision is made using the information from the chance node.

Do you know if there's an easy way to calculate that in GeNIe, without adding a prior decision node, or do I need to do a side calculation?

Thanks for your help.
marek [BayesFusion]
Site Admin
Posts: 430
Joined: Tue Dec 11, 2007 4:24 pm

Post by marek [BayesFusion] »

regnier wrote:This doesn't actually give me what I want. What I would like is the expected value of the optimal decision at the decision node, when that decision is made using the information from the chance node.

Do you know if there's an easy way to calculate that in GeNIe, without adding a prior decision node, or do I need to do a side calculation?
I'm a little confused about your original question now. It seems to me that GeNIe will calculate what you want without any tricks. The answer is a table indexed by the chance node that will be observed before the decision node. Please note that the expected value depends on what is going to be observed and you have not observed anything. If you have observed a state, then the table at the decision node will not be indexed, of course, and you get straight what you want. Does this help?
Cheers,

Marek
Post Reply