forum.hugin.com

User Forums => HUGIN GUI Discussion => Topic started by: Karol Strzelinski on May 18, 2009, 19:52:31

Title: Logarithmic score based model performance indicators
Post by: Karol Strzelinski on May 18, 2009, 19:52:31
Hello,

I am new Hugin user and Bayesian Networks are also quite fresh topic for me. I am experimenting with models for operational risk purpose and I would like to assess them, to choose the best one. I have noticed AIC, BIC and loglikelihood criteria in Analysis Wizard,  but I am wondering if it is possible to perform somehow analysis using logarithmic score based model performance indicators like node monitors, global monitors, etc. (more on these indicators in R.G. Cowel,  Probabilistic Networks and Expert Systems).
I will appreciate for any help in this matter.
Title: Re: Logarithmic score based model performance indicators
Post by: Anders L Madsen on June 15, 2009, 21:32:35
It is not currently possible to analyse model performance using logarithmic score based model performance indicators. Please provide a more detailed description of the type of performance indicators you would be interested in.

I apologize for this belated reply.