Palisade Knowledge Base

HomeTroubleshooting@RISK for Excel: Simulation"No values to graph" Message / All Errors in Simulation Data

4.45. "No values to graph" Message / All Errors in Simulation Data

Applies to: @RISK for Excel 5.5.0 and newer

When I run my model, Browse Results shows "There are no data values to graph" for my outputs. If I look in the Simulation Data window, it shows "Error" in every row. What is wrong?

There could be two causes. If this is a legacy model from @RISK 5.0 or later, see My model worked just fine in @RISK 5.0 (or 4.5), below.

If you created this model for @RISK 5.5 or later, it's more likely that you have a logic error in your model. You can use "shoeprint mode" or "footprint mode" to find the error. Open the Simulation Data window (small x-subscript-i icon in the Results section of the ribbon), and then proceed as in the article Some Iterations Show Error in Data Window. What Can I Do?

My situation is similar, but Simulation Data shows "Filtered" on every row, not "Error".

You have a filter defined, and its specifications cause every value of this output to be screened out.

Look at the Define Filters button, near the middle of the @RISK ribbon. If a filter is active, the button has a colored background. Click the button, and if you wish you can edit the filter settings or simply remove the check mark from the box next to Enable Filters for Simulation Results.

If you're editing filters, you may wonder about the Type column. A Standard filter affects only that particular input or output. An Iteration filter uses the values or percentiles of that input or output as a criterion, but when the Iteration filter rules out any particular iteration of this input or output, @RISK also rules out the same iteration of all other inputs and outputs.

My model worked just fine in @RISK 5.0 (or 4.5). But when I run it in 5.5 or later, I get "There are no data values to graph" for my outputs, or the Simulation Data window shows "Error" in every row. What is wrong?

Are you using RiskPercentile, RiskMean, or other statistic functions? These behave differently in 5.5 and later from how they behaved in 5.0 and earlier (including 4.5).

In the earlier versions, @RISK statistic functions like RiskMean and RiskPercentile were evaluated during every iteration. This created three problems. First, for iterations 1 through N-1, the values were wrong because they were computed only on the iterations available so far. Second, and related, the values changed as each additional iteration was run, which was generally not what users wanted or expected. Third, computing all these extra values slowed down the simulation.

Beginning with 5.5, by default the statistic functions are computed only once, at the end of the simulation. This can make a dramatic speed improvement, and it also makes logical sense because the means and percentiles and such aren't known until all iterations have been computed. However, in a few models this reveals a problem that existed all along, but that users were unaware of: trying to use a value (such as a simulated percentile or mean) before the final value was available.

If your model relies on the old behavior, and you actually want to recompute the statistics in each iteration, you can make @RISK behave like the older versions. In Simulation Settings, on the Sampling tab, change "Update Statistic Functions" from "at the end of each simulation" to "each iteration". If you make that change, so that @RISK 5.5,–7.x behaves like 4.5 and 5.0, please bear in mind that this may let your model run at the expense of papering over a real problem. Any formula that refers to a percentile, a mean or another simulated statistic function would get a different value at each iteration. Especially in the earlier iterations, it would be quite volatile.

To get statistics of input distributions, the better method is to use RiskTheoPercentile or RiskTheoMean or similar, rather than RiskPercentile or RiskMean. If you're interested in statistics of an input distribution, a better procedure is to use the "Theo" functions. See Statistics for an Input Distribution. But if you're taking percentiles or means of an output, they cannot be known in advance and your solution is either to change Simulation Settings as mentioned above, or to re-examine the logic of your model.

Last edited: 2017-02-23

This page was: Helpful | Not Helpful