Box 10.9 Step 2 detailed calculations

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Box 10.9 Step 2 detailed calculations

Pablo
I cannot manage to reproduce step 2 results in Agena. Can you develop a little bit more the trick that you referred at the top of page 331?

Does Agena have any programming procedure to calculate bayesian scores in bayes nets with many more nodes?

I also appreciate if we can know about a paper where to read about bayesian scores.

Great book!!

Many thanks.
Reply | Threaded
Open this post in threaded view
|

Re: Box 10.9 Step 2 detailed calculations

norman@eecs.qmul.ac.uk
Administrator
Pablo,

To understand the 'manual' calculations you need to look at the AgenaRisk model (but there are some minor discrecpencies between the values presented in the book). So look at the submodel "H1 independent". To find the value of P(r=19, n=151 | H1) what you have to do is look at the graph of the node labelled "19 deaths in 151". If you display this graph in its own window and then slide the right hand vertical toolbar to display the statistics you will see that it tells you exactly what the marginal probability of the value 19 is. In the latest build of AgenaRisk the value is 0.072 (as opposed to 0.071 written in the book).  Similarly if you look at the statistics for the node "11 deaths in 63" you will see the probablity for the value 11 is 0.049 (as stated in the book) etc

The joint probability is then just the product of the individual probabilities.

Please see also previous postings about this.

Regarding:

>Does Agena have any programming procedure to calculate bayesian scores in bayes nets with many more nodes?

There is a programming API that comes with the enterprise version of Agena. We have been discussing the possibility of making a version of the API available for free - so watch for announcements on the Agena website.

>I also appreciate if we can know about a paper where to read about bayesian scores.

Do you mean the Bayes factor scores?
In case you haven't seen them there have been a number of previous threads about Bayes scores.

Norman Fenton