Frequently Asked Questions

Why am I getting messages telling me that 'inconsistent evidence' has been entered on one or more nodes?

The method of probability calculation carried out in AgenaRisk aims to make the overall  model consistent with the assumptions made and the observations entered. The model will be inconsistent if you:

  • Select a state value as an observation when the probability of observing that state is actually zero

  • Entering observations on two state values on different nodes each of which have non-zero probabilities but which when calculated actually have zero joint probability of occurring together.

  • Create a risk node containing a probability assignment that depends on another node that contradicts this assumption.


In these cases AgenaRisk will attempt to diagnose the cause of the problem and will display the issue in an error message. AgenaRisk will then “roll-back” the calculation made and display the last consistent state of the risk model. If this happens you will have to remedy the model by either withdrawing contradictory or impossible observations or amending the probabilities accordingly.

When using simulation nodes you may also get an inconsistent evidence message in situations where theoretically the probability of the observation is non-zero. This is because although a simulation node with an expression like, for example, the Normal distribution, may in theory have non-zero probability values over an infinite range there are ranges whose probability is so close to zero that it is treated as zero. To understand this, create a new simulation node. By default the NPT is a Normal distribution with mean 0 and variance 1000000. Change the variance to 1000 and calculate.  Although in theory arbitrarily high or low values are possible for this distribution (such as a value within the range 100000000 and 100000000.1) such probability values are so small that if you entered an observation within such a range you will get an ‘inconsistent evidence’ error message. In fact, almost all of the non-zero probability mass (99.9% of it) lies within –30 and 30.  If you try to enter a value above, say 300 you will trigger the error message.


My model was previously working OK, but now it does not calculate properly. What is going on?

The most frequent reason that a previously working model will not run properly in the new version is because of the new simulation convergence settings. In versions of AgenaRisk prior to version 6.0 the default simulation convergence was 1.0E-6.  Before running any old model in the new version you should change this setting to the new default of 1.0E-2.

Sometimes (especially when using older legacy models) a model can become corrupted so that a previously working model starts to behave in unexpected ways and even fails to calculate.  A common symptom of a model that has become corrupted is a very large cmp file. This can sometimes happen if the user had cut and pasted a jpeg or similar into a text field (we have now designed out this failure mode). The use of foreign language characters (such as umlauts and circumflexes) has also led to instances of corrupted models.

The steps to determine and fix corruption are:

  1. change .cmp file name to .zip (to enable windows to view it as a compressed archive)

  2. open up this .zip file and find the file with .mod

  3. if the .mod file is >1MB in size then it is most likely corrupt (windows or the zip file app. should report the uncompressed size of this file).

  4. copy the .mod file from the zip archive and change it to a .txt extension

  5. open it in wordpad or similar.

  6. scroll down and if you see many strange characters together then it is corrupt

  7. find the node name above these strange characters

  8. go back to original cmp file and open in Agenarisk

  9. find the risk object containing that identified node and delete it (keep a backup!)

  10. save model and then reopen and recreate risk object you deleted - from scratch.

Memory issues: Why does the application sometimes hang in a way that it cannot be recovered and is there anything I can do stop this happening?

Large models with many simulation nodes and with high accuracy simulation settings require a lot of memory to run. You may be able to increase the amount of memory available depending on the version of AgenaRisk installed and on your machine's processor. AgenaRisk has a choice of 64-bit or 32-bit installer. The 32-bit version limits memory size to 1 GB but with the 64-bit version the only limit is the amount of RAM in your machine. In this case you just need to edit the default settings in  text file named "AgenaRisk.vmoptions" that you will find in the installation directory (this is  typically located here: C:\Program Files\AgenaRisk). So, for 1 GB memory we set -Xmx1024m and for 3GB we set -Xmx3072m (this is the default). There is a similar setting for stack size, which you might need to open very large models.


Why does the application seem to hang sometimes when I enter an observation in the risk table view or in the risk map view?

It is almost certain that you have the auto-calculate button set on. When it is set on AgenaRisk will automatically run the whole model every time you enter a single observation and that is what is causing the delay.  For most models it is recommended that you do NOT have auto-calculate set on. 


Why is the graph monitor not displaying when I double-click on a node? Similarly when I right click on a node nothing happens.

This is almost certainly because you unchecked "Filled" in the node Properties--> Appearance. Doing this actually makes the node 'transparent' except at its edge. So, if you really want it to be transparent for aesthetic reasons then you can still double-click on the edge to display the graph monitor or right click on the edge to show the properties.  However, if the reason for unchecking "Filled" was to make the node colour white (which is the default, but may have been changed) then the way to do this is to simply select the colour white with "Filled" still checked.