Can I make Prism mark certain bars or symbols with asterisks (to indicate statistical significance)?

Automatically add lines or brackets with stars or P values from appropriate analyses

Prism 9 introduced the ability to automatically add results from specific analyses containing pairwise comparisons onto your graphs. After performing an appropriate analysis, switch to the graph where the data is plotted and click the "Add or format pairwise comparisons" button in the Draw section of the Toolbar. To customize the appearance of the added results (e.g. to change the line thickness or style, to display actual P values instead of "star" summaries, etc.) or to select which specific comparisons are displayed on the graph, click the "Add or format pairwise comparisons" button again, and use the "Format Pairwise Comparisons" dialog to specify the desired look of the graph.

To learn more about the Pairwise comparisons feature (and which analyses are supported), check out this page in the Prism User Guide.

Quickly add lines or brackets with centered text to indicate significance

Prism 8 introduced a new drawing feature that allows you to draw lines or brackets that include a centered text box. When you draw the line/bracket, the text box is automatically added and includes a number of default selections to indicate significance that you can simply click (ns, *, **, ***, #, ##, ###). Note that any custom text can also be entered into this centered text box. When the line or bracket is moved, the text automatically adjusts appropriately.

What are some alternatives to the automatically added pairwise comparisons?

Forget about asterisks. Plot confidence intervals instead. 

Starting with Prism 6, you can make a new graph for the confidence intervals for the differences between means (in the Options tab for one- or two-way ANOVA when you choose multiple comparisons tests). I think this lets you show the results much more clearly than a bar graph with asterisks. 

In Prism 9, this is taken one step further for t tests. In this case, you can create an Estimation Plot, showing the raw data for both groups, the difference of the means between the two groups, and the 95% confidence interval of the difference between means on a single graph. This is a great way to summarize the results of this test.

Write the asterisks manually

• Use the "Lines with centered text" feature described previously.

• Copy asterisks from the results page and paste onto the graph. Then move into position. Prism will snap them into place over a bar or between two bars, which should make it easy. These pasted asterisks are linked to the analysis so will update if you edit or replace the data. 

• Use the text tool to write P values or asterisks, and then position. These, of course, will not be linked to the analysis. 

With either of these methods, notice that the draw menu lets you create an up-over-down bracket to indicate what the P value or asterisk is comparing.

Sort-of-kind-of automate

If you frequently use the same kind of graph, you could make a template using a work-around. This technique applies to marking both point symbols and bars, but if you want to mark bars, you will need to make a "pseudo" bar graph as shown in this article.

Mark the bars or symbols by adding a second set of points:

1. From the data table, do a transformation using Y=Y+K, where K is the distance above the top of the bar that you would like to plot an asterisk.
2. Add the resulting data set(s), which will be on a linked results sheet, to your graph, and change the symbol for that data set to " * ".
3. Store the template.

All your bars will be marked, so each time you use the template, you must remove asterisks for the bars you don't want marked. Enter the new data, then go to the results sheet for the transform and exclude (don't delete) the Y+K values for the bars that are not to be marked.

If you have error bars and the asterisks would collide with them, you could add an additional transformation of the X values (X=X+K) to displace the asterisks horizontally (as well as vertically).