Creating Interval Node for Normal Distribution

[sandra]

Just a quick easy question:

I am using the Java API to create a BN and I want to create an interval node which discretises a normal distribution with say mean 0.25 and variance 0.15.

I got as far as :

IntervalDCNode intNode = new IntervalDCNode (domain);
intNode.setNumberOfStates(10);

and then didn't know how to set the distribution and ranges that I want the intNode node to have.

I also don't know how to assign a normal distribution to a continuous node:

ContinuousChanceNode contNode = new ContinuousChanceNode(domain);

then what?? I expected there to be a method such as setDistribution(normal, mean, variance) - wishful thinking!

I looked at the Java examples and API documentation, but it is not clear to me how I create it.
The constructor needs two Expression objects, one for the mean & one for the variance, but don't know how to create the Expression objects to represent the mean & variance

Thanks.

[Anders L Madsen]

Hi Sandra,

I got as far as :

Code: [Select]

IntervalDCNode intNode = new IntervalDCNode (domain);
intNode.setNumberOfStates(10);

and then didn't know how to set the distribution and ranges that I want the intNode node to have.

You should use the method

Code: [Select]

setStateValue to define the intervals.  This discussion item describes how to create an expression that sums two numbered nodes:
http://forum.hugin.com/index.php?topic=111.0
The principle is the same for intervals nodes. Notice that HUGIN does not know how to identify the intervals, i.e., perform the discretization.

I also don't know how to assign a normal distribution to a continuous node:

Code: [Select]

ContinuousChanceNode contNode = new ContinuousChanceNode(domain);

then what??

By definition a continuous node has a Normal distribution. Thus, you should use

Code: [Select]

setAlpha
setBeta
setGamma on ContinuousChanceNode to set the parameters of the distribution.

Hope this helps. If not, ask again:-)

[sandra]

Great, thanks Anders.
Too easy!
Thanks for your quick response.
Cheers,
Sandra