Performance of performing multiple comparisons is low when input data contain formatted (bold, italic, etc.) text

The Problem

Prism will occasionally run slowly when calculating multiple comparisons following two-way ANOVA with the option "Compare cell means regardless of rows and columns" due to a huge number of groups (rows and columns) in the input data with formatted text (bold, italic, etc.). In some cases, it may even seem that Prism is frozen.

The Reason

When Prism calculates multiple comparisons, it generates a sheet of results that include the names of each group being compared in every comparison. When using the option "Compare cell means regardless of rows and columns", the total number of multiple comparisons being compared can quickly become extremely large. Consider an input data table for a two-way ANOVA with M rows and N columns. The total number of unique groups in the data can be given by K = M x N. The number of multiple comparisons possible from this table is then calculated as ((K)*(K-1))/2. For small values of M and N, this is normally a reasonable number. For example, when M=4 and N=3, K=4*3=12, and the total number of multiple comparisons is 66. However, consider an example in which M=50 and N=30. Now, K=50*30=1500 and the total number of multiple comparisons is over one million. Under these conditions, Prism is still able to perform the calculation, it simply takes a while to complete. In general, the higher the dimension of your input data, the longer it will take Prism to complete these calculations.

When group labels (column and row labels) contain text with formatting, this challenge is compounded. When generating the results of multiple comparisons, Prism lists the information for each multiple comparison along with its (formatted) title. For example, "[Col 1]:[Row 1] vs. [Col 1]:[Row 2]". Storing the text formatting styles for these results is suboptimal. Given a huge data table with formatted row/column titles, it takes Prism a long time to generate and render each of these formatted titles for the results sheets. Moreover, this issue may be encountered when opening or autosaving the file. We will consider methods to improve this behavior in future versions of Prism.

Workarounds

You may try one or a few of the following workarounds:

• Remove formatting from titles in the input data table
• Reconsider using the "Compare cell means regardless of rows and columns" method for huge data tables (are these really the comparisons you planned prior to conducting the experiment?)
• Consider disabling auto-backups in Prism preferences and saving the files manually

If you have a file with a huge number of formatted titles and are having trouble opening the file, you may contact our support team and we will do our best to help fix the file.