Graph Builder: Combining Measurement Data & Model Predictions with CIs

[#1]

Hello everyone,

I'm working on creating a combined graph in JMP's Graph Builder, and I'm encountering a few challenges. My goal is to visualize both experimental measurements and predictive model outputs, including their respective confidence intervals, on a single plot.

Here's a breakdown of my data and the issues I'm facing:

Data Description:

• Measurements.jmp: Contains X and Y data, represented as mean values with standard deviations.

  • Data is categorized by two groups (0 and 1), which I intend to represent with different colors via the overlay feature.

  • Different dates are also present, which I'd like to show with distinct markers.

• Models.jmp: Contains the parameters for a linear function ($y=a+b\cdot x$) for each group. These parameters (a and b), along with their confidence intervals, were derived from the Fit Model platform.

Challenges and Questions:

1. X-axis Error Bars: I've successfully added standard deviation error bars for the Y values in Graph Builder. However, I haven't found a way to display standard deviations for the X values. Is this functionality available in JMP's Graph Builder?

2. Group-Specific Confidence Intervals for Models: Each linear model (for Group 0 and Group 1) has its own unique set of confidence intervals. While I can plot both model functions, I'm currently only able to display a single, uniform confidence interval, which isn't appropriate for my data. How can I plot the distinct confidence intervals for each group's model on the same graph?

3. Combining Datasets with Different X-values: My Measurements.jmp and Models.jmp datasets have different  X value ranges. I'm struggling to combine these datasets for plotting in Graph Builder, as it seems to primarily support plotting two different Y datasets against a shared X axis. How can I integrate data with disparate X ranges onto a single graph?

4. Representing Confidence Intervals as Semi-Transparent Areas: Instead of discrete bands or bars, is it possible to represent the confidence intervals as a semi-transparent shaded area on the plot?

I'm open to solutions involving JMP scripting, but a solution using the Graph Builder GUI would be highly appreciated if possible.

Thank you in advance for your assistance!

[#2]

1. You can try to swap your response axis (you might have to add multiple point plots)

2. Stack your data so you can do the separation using Overlay or some grouping zone (X Group, Y Group, Page). 

After stacking the data you most likely want to rename your columns, remove extra columns and add separate column for model number

3. You can concatenate your tables in such a way that they have same name for X-axis (and Y axis) and then plot them

Or you can try to manage with multiple columns on both axis

4. Add line chart on top of your points chart and use that for the error interval

[#3]

Dear Jarmo,
thank you very much for your quick reply.

Your suggestion to use stacking and add a line chart was helpful and allowed me to display distinct confidence intervals for the model's functions as a shaded area. Thank you!

However, I am still struggling with combining both datasets effectively within a single graph.

3. Concatenating the datasets prevents me from manipulating them independently, which is needed for ensuring the measurement data points align correctly with their respective model lines. Conversely, when keeping the data in multiple columns, I struggle with properly setting different error bars and confidence intervals. Specifically, dragging and dropping four columns into the "Interval" field seems to only register three, preventing me from applying all necessary error bars/bands.
I observed in your example that you managed to obtain separate "Points" left menus for the measurement data points and the model functions. While I can achieve this through scripting only, I am curious if there is a method to accomplish this using only the JMP GUI.

1. I am still attempting to determine how to add error bars to both the X and Y data points simultaneously. As you demonstrated, it's possible to swap responses to define an error bar for either X or Y, but not for both at once.

I would greatly appreciate any further guidance.

[#4]

You can right click on the graph builder and add multiples of the same plot type

You can also drag and drop plot type to the graph 

You can add double points to your graph, one for X and one for Y response axis and then set the error bars