The problem

Estimation plots generated from paired t tests in Prism 9 or newer incorrectly include some additional connecting lines when opened using Prims 8 or older versions of Prism.

What you see in Prism 9 or newer (correct)

What you see in Prism 8 or earlier (incorrect)

More Details

Estimation plots were introduced in Prism 9, and serve as a great way to summarize the results of tests dealing with comparisons of means such as unpaired or paired t tests. When performing either an unpaired or paired t test, Prism will generate an estimation plot by default. These graphs plot the raw data for each group as well as the difference of the group means and the 95% confidence interval of that difference.

When running a paired t test, it is assumed that the values in each group are paired or matched with values in the other group. This pairing is shown on the graph by connecting the points representing paired values with lines between the two groups. In Prism 9, these connecting lines are correctly shown only for the points in the raw data. The connecting lines do not extend to the summary displaying the difference between values, the mean difference, and the 95% CI of the difference.

However, if an estimation plot is saved in Prism 9 (or newer) and opened using Prism 8 (or older), the connecting lines will incorrectly be shown for all groups on the graph.

Are there any workarounds?

There are a few ways to remove the unexpected lines from an estimation plot opened with Prism 8 or earlier. Steps describing two of the easiest ways to do this are given in the sections below

Change the Style/Appearance of one data set

• Open the Format Graph dialog by double clicking on the graph
• Select the second-to-last group from the Data Set dropdown menu (the group that is incorrectly connected to the summary values on the graph). In the example below, this group is "Treated"
• In the "Style" section of the format graph dialog, select "Aligned dot plot" in the Appearance dropdown menu
• To hide the summary (mean) line for this data set, set the transparency of the line to completely transparent in the "Lines" section of the Format Graph dialog
  • For Mac: click the "Color" box and move the slider for "Opacity" to zero. Close the colors dialog
  • For Windows: click the "Color" dropdown menu and select "More colors & transparency...". Move the transparency slider to 100% and click "OK"
  Click "Apply" or "OK" on the Format Graph dialog

Incorrect Graph Appearance

Select the correct Data Set in the Format Graph dialog

Select "Aligned dot plot" in the Appearance dropdown menu

Set the "Lines" color to be completely transparent

For Mac

Open the Colors dialog by clicking on the "Color" box in the "Lines" section of the Format Graph dialog

Set the "Opacity" slider to 0%

For Windows

Click the Colors dropdown menu

Select "more colors & transparency"

Set the "Transparency" slider to 100%

Click "OK" or "Apply" to get the expected Estimation Plot appearance

Change the color/transparency of the connecting lines for one data set

Instead of changing the graph style, you can instead just change the visibility of the unwanted lines.

• Right click on any symbol in the second data set (the group that is incorrectly connected to the summary values on the graph)

For Mac

• In the contextual menu, go to "Format Entire Data Set" and choose "Line/Curve Color & Transparency"
• In the dialog that appears, move the "Opacity" slider to 0% and click OK

For Windows

• In the contextual menu, go to "Format Entire Data Set" and choose "Line/Curve Color", then choose "More colors and transparency..."
• Slide the "Transparency" slider as high as it will go (95%) and click OK

Note that because this method does not allow setting transparency to 100%, there may still be very slightly visible lines on the graph. Because of this, we recommend using the earlier method using the Format Graph dialog.