## AgenaRisk Professional Version 6.1 released 11 February 2014We are pleased to announce the release of Version 6.1 of AgenaRisk - a powerful but intuitive tool for modelling, analysing and predicting risk using risk maps, otherwise known as Bayesian networks. Its simulation capabilities go beyond Monte Carlo methods.
1.Entropy error values are now displayed alongside other statistics in risk graphs. 2.Calculation time and memory are displayed at bottom of AgenaRisk screen. 3.Simulation convergence now governed by entropy error rather than number of iterations. Model setting default is 0.001 and this is recommended for most models. 4.Mixture distributions involving constants and arithmetical functions of constants now handled properly 5.Old default simulation convergence scheme changed slightly. Model level setting is default for all nodes not individually set. 6.Statistical distribution bounds tidied up so that you no longer need to worry about specifying valid interval state ranges (in fact you don’t need to bother with these at all now). The new default state values that are set for all types of simulation nodes work for all types of ranges and distributions. There is no no need to change them. 7.Better reporting of inconsistencies and “out of memory” conditions. 8.Various warnings now triggered if a) model has a higher than default simulation accuracy setting b) too many risk graphs are open demanding memory. 9.Extreme value distribution improved to deal with maxima and minima. 10.Input nodes can now be simulation nodes. 11.Models from book added to the /Examples directory for easy access. Also several new example models added and some old models tidied up and improved. 12.Output log tidied up and it now flags singularities and potential overflows (though these are now correctly handled). 13.By default statistics are displayed alongside risk graphs when a risk graph is displayed in its own window.
1.Zero compression 1 - After simulation this removes states with zero probability mass from ends of marginal distributions, making for more accurate results. 2.Zero compression 2 - During simulation this removes states with zero probability mass in middle of marginal distributions, making for faster results on multi modal and mixture models. 3.Truncation - After simulation removes intervals with log entropy error less than average in order to remove unnecessary “tails”. 4.{-inf, +inf} states no longer appear at end of marginal distributions. These used to make summary statistics inaccurate in some cases. 5.Multiple interval splits per iteration. Results in (generally) few iterations and speed up on models with large cluster sizes. 6.For simulation nodes you should never need to enter any intervals to define the domain for the NPT. AgenaRisk should now do it automatically. 7.NPT generation and expression handling now done during calculation, when necessary, and never during model editing. 8.The simulation algorithm is now much better at expanding and contracting bounds on nodes on the fly, depending on the current state of the model.
1.Input and output node parameter passing bugs fixed. 2.CSV scenario import/export issues fixed 3.X and Y axis displays full range for distribution. Before if you specified an X-axis range A to B to display on the risk graph it would only display the range A to B-1. 4.Better parsing of expressions to determine if reserved words are used and whether deterministic and Boolean expressions are correct. 5.No longer removes small probability intervals during simulation since this can lead to cycling 6.Input node state values revert to original values after calculation. 7.Bug where constants were not handled properly when declared on Boolean and Labelled nodes fixed. 8.Bug where notes fields would become corrupted now fixed. 9.Numerous small annoying bugs fixed in simulation, input/output parameter passing and the handling of node types. |

11 Main Street Caldecote Cambridge CB23 7NU, UK Tel: + 44 (0) 1223 263880 Fax: +44 (0) 870 1318460 www.AgenaRisk.com | sales@agena.co.uk © 2010 Agena Ltd. | All Rights Reserved. | Privacy Policy |