What inspired this wish list request? In using the Variability Charts frequently, there are numerous issues I have in being able to customize the plots flexibly and dynamically. Often, I must settle for what's available since the customizations aren't ideal. The customizations are also hidden in many cases, limiting their potential for more casual users of the platform.
What is the improvement you would like to see? Below is an itemized list of pain points that can be improved to enhance the user experience when using the platform.
Why is this idea important? Variability Charts still get used a lot but have had less development work. They work well with many cases with nesting and many of the visually appealing features are simpler in this platform. Nesting works better and more naturally in Variability Charts than in Graph Builder.
1. The top chart with the raw data is clunky when customizing the Group Means and Connect Cell Means. You need to right-click then Customize in the white space of the graph, which is hidden for most people. Once the Group Means is found, it doesn’t really work since the reference is actually a purple line that doesn’t match what’s in the graph. It would be best if the legend in the top right could be interactive as it does in Graph Builder (Overlay, Color) or Fit Y by X (Compare Densities graph).
2. The default Std Dev chart has no separation panels. If there are at least 3 X variables and I add the Group Means for Std Dev, the grids will enter for all X variables. The grids should match the X variables shown. A snippet of this output is shown below (only Operator was selected).
3. The color of the Group Std Dev cannot be changed. There is also no customization if multiple X variables are chosen for colors, line styles, or line widths.
4. The drag-and-drop feature of the X variables in the nesting of the plots is very useful and flexible. An additional feature to consider would be to remove an X variable (likely because it seems unimportant) rather than rebuild a new plot with the variable removed. Another similar feature would be a way to add an additional X variable as well.
5. Adding a Row Legend can be an effective enhancement to visualizations, but it works best the same way the Overlay/Color functions in Graph Builder work. When applied in Variability Chart (and other platforms) it is applied globally, which is likely undesirable. If you customize the colors in the Row Legend, but then save the script, there is an error, and it will revert back to the original Row Legend values.
6. A smaller issue I have is with the variance components. The 20/40/60/80 % labels for the graph seem like a good idea, but they tend to just be smushed together and not really add clarity. You could argue that removing the label would be similar to how the Effect Summary displays the p-values in the Fit Model platform where you have hashmarks representing 0.1/0.01/0.001, etc.
