Introducing the Final Value Result: Soon Available in GoldSim 14

Posted by Rick Kossik

We have been working for the last several years on the next major GoldSim release (GoldSim 14).  We expect to release this by October 15. All users with active maintenance will receive an email at that time with instructions for upgrading.

GoldSim 14 contains a large number of changes and updates.  In this blog post, I briefly discuss one of the most noticeable new features that you will see (a new type of result display). When we release GoldSim 14, we'll post another article to the blog (as well as send out a newsletter) summarizing all of the new features and updates.

The Final Value Result

GoldSim 14 introduces a powerful new result display option referred to as a Final Value Result. Final Value results are used to carry out side-by-side comparisons of the final values of outputs for different types of simulations.  In order to use them, you need to carefully specify what you wish to compare and how you want to display that comparison. Depending on how your model has been run (e.g., multiple realizations, multiple scenarios, multiple Capture Times) and the outputs you have specified to compare (e.g., one output, multiple outputs, vector results, matrix results), there can be a large variety of ways to display the comparison.

One of the simplest Final Value results you might want to display is simply the comparison of multiple outputs for a deterministic run or a single realization.  For such a result, there are three meaningful ways that you can choose to display this in a Final Value result. This is a column chart:

This a bar chart:

This is a pie chart:

You can also this same data in tabular form, either as rows or columns of values (below it displayed as rows):

We can make this “one-dimensional” data set “two-dimensional” by running a Monte Carlo simulation. In this case, the final values we would want to look at would likely be statistics (e.g., the mean or some percentile). In this case, the tabular form of the data would look like this (as in the previous case, the rows and columns can be reversed):

Now that we have “two-dimensional” data, our charts can become a bit more interesting and we can explore and compare the results in different ways.  For example, we could produce a clustered column chart like this: 

This chart displays three groups (clusters) of columns.  Each group represents a different output, and each column represents a different statistic for that group.  The Final Value result, however, can also “flip” how this information is displayed:

In this case, each group of columns represents a different statistic, and each column represents a different output.

These two displays are in the form of column charts, but we also display bar chart versions of these same results. For example, here is the clustered column chart above displayed as a clustered bar chart:

We can also display “two-dimensional” pie charts:

In this case, each pie represents a different statistic, with each pie slice representing a different output. Alternatively, however, we could specify that each pie represents a different output, and each pie slice represents a different statistic.

Finally, having a “two-dimensional” set of data allows us to make use of stacked bar and stacked column charts.  That is, rather than presenting the data using groups of columns (or bars), we can present it using stacked columns (or bars):

Stacked bar and column charts can also be displayed as percentages (of all the items in the “stack”) rather than values.  Here is what the stacked column chart above would look like if displayed as a 100% stacked column chart:

Such a chart is similar to a pie chart (in that all of the sections add to 100%).

 Running your model in different ways can add additional “dimensions” (beyond two) to the set of data that you wish to display.  For example, we could have run this particular model for two different scenarios. We would then have three “dimensions” to our data set (outputs, statistics, scenarios). Charts themselves can not extend beyond “two-dimensions” (like those shown above), so when we have three “dimensions” in the set of results, we need to select a single value for one of the dimensions (referred to as the Layer) when displaying the chart.  But having three dimensions provides great flexibility in how we display the data.

For example, we could choose to display a column chart showing all outputs and all statistics for a particular scenario (in two different ways):

Alternatively, we could choose to display a column chart showing all outputs and all scenarios for a particular statistic (in two different ways):

Finally, we could choose to display a column chart showing all scenarios and all statistics for a particular output (in two different ways):

In this case, we produced six different column charts displaying these results in different ways (which allows you to emphasize different aspects of the results).  Note that we could have also produced equivalent stacked column charts, bar charts, stacked bar carts, and pie charts (six different charts for each type of display).

The tabular view of this data would look like this (and could also be displayed in multiple ways, by switching what is displayed in the rows and columns):

In the case of tables, unlike charts, the “third dimension” can be fully displayed (GoldSim simply increases the number of columns).

The simple examples provided here do not cover all of the combinations of results that you can display (e.g., Final Value results can also display arrays and results for different Capture Times), but what you should conclude from this overview is the following:

• The Final Value result is very powerful and flexible, and allows you to produce a variety of charts and tables to display results at the end of the simulation (Final Values) or at defined Capture Times.
• At their most complex level, charts can display up to two “dimensions” of results from higher dimensional data sets (e.g., multiple outputs and multiple statistics for a single scenario; multiple outputs and multiple scenarios for a single statistic, multiple statistics and multiple scenarios for single output).
• Tables can display the full data set selected (by increasing the number of columns).
• The various charts and tables can be displayed and rearranged in a wide variety of ways, giving you great flexibility in order to emphasize specific aspects of the results.