---
title: "Scatter plots"
canonical: "https://modelassist.epixanalytics.com/space/EA/26575484/Scatter%20plots"
format: markdown
---
By plotting the generated values for an input together paired with the output values in a scatter plot one can get perhaps the best understanding of the effect of the input on the output value. Plotting generated values for two outputs is also commonly done: for example, plotting a project's duration against its total cost. Scatter plots are not directly available in Crystal Ball but are easy to produce by exporting the simulation data from Crystal Ball at the end of a simulation into Excel (select "Run" >> "Extract Data" >> Forecast Values" >> "OK"), provided you have told Crystal Ball to collect the input generated values for each input distribution of interest.

 

*Note:* While in the Crystal Ball versions before 7.0, one had to select an assumption as a forecast to collect its randomly generated values, Crystal Ball 7.0+ offers the option to automatically collect all forecasts on every assumption.

 

It takes a little effort to generate these scatter plots, so we recommend that you perform a rough sensitivity analysis to help you determine which of a model's input distributions are most affecting the output first.

 

![image](media://3a9db8d0-fc0d-4a93-99b0-7240b97adcb1)

 

This plots shows 3000 points, which is enough to get across any relationship but not too many to block out central areas. The chart tells the story that the model predicts increasing advertising expenditure will increase sales – up to a point.

Since this is an Excel plot we can add a few useful refinements. For example, we could show scenarios above and below a certain advertising budget:

 

![image](media://c8ffb14d-bc78-4225-a4d8-f5f044bb362c)

 

We could also perform some statistical analysis of the two sub-sets, like a regression analysis (here's how in an Excel chart):

 

![image](media://3357b069-44a7-4c09-8cda-c09dc61e655b)

 

The equations of the fitted lines show that you are getting about three times more return for your advertising dollar below $150k than above (0.0348/0.0132 ≈ 2.6).

It is also possible, though exceptionally tedious, to plot scatter plot matrices in Excel to show the inter-relationship of several variables. Much better is to export the generated values to a statistical package like SPSS.

 

 

 

---