Quantcast

Handling many data points (goodness of fit)

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Handling many data points (goodness of fit)

norman@eecs.qmul.ac.uk
Administrator
A reader has asked the folliowing in relation to Example 10.8 (page 325) for 'goodness of fit'.:

"I managed to replicate the network from this example in AgenaRisk however I would like to apply this to my own data. In the example there is an observation node for each data point and this is used to infer the likely distribution to give the data (I believe). However I have many data points and it is not feasible to generate an observation node per data point. Therefore I would like to know if there is a way in which I can make many observations with just several nodes? The data is continuous and I am trying to fit a distibution to the histogram that I have produced. I have done this in Excel based on maximum likelihood but would like to do the same using Bayesian methods (I have five or so candidate distributions). Any help would be much appreciated."

Currently, the only way to do the full Bayesian analysis  in AgenaRisk that the reader requires for this type of example (without creating a node for each datapoint)  is to do a little programming using the AgenaRisk API which comes with the Enterprise version. In a future release Agena is intending to incorporate some of this kind of more automated parameter learning into the desktop version.

Loading...