Fixed bug: Compare Observed and Expected distributions with Prism 6.0a Mac with three or more groups

Prism 6 introduces a new analysis to compare an observed categorical distribution with distribution expected by theory. With three or more groups, this is done by the chii-square test. With only two categories, Prism offers a choice of the chi-square test or the binomial test (which we recommend). This works fine in Prism 6.00 Windows.

With 6.0a Mac, Prism always defaults to computing the binomial test, even when there are three or more categories, a situation where the binomial test is meaningless. The choice of binomial vs. chi-square is gray (unavailable) with three or more tests, as it should be since the binomial choice makes no sense with three or more categories. But if the choice is set to binomial test, which is the default, Prism 6.0a Mac does the wrong calculations. It labels the results as "binomial test" and reports separate one- and two-tailed P values, which is meaningless when there are three or more categories. So the results are obviously meaningless. 

Simple workaround:

  1. Enter data onto only two rows. Or enter data on more than two rows, but exclude all but two rows.
  2. Run the analysis to compare observed and expected distribution. Choose the chi-square, not the binomial, test. Enter any values as expected values, as you won't care about the results. But the expected percentages must add to 100% or the expected values must add to the sum of the observed.
  3. Go back to the data table and enter the rest of the rows. Or unexclude them.
  4. Go to the results sheet. Change parameters, and enter the expected numbers or percentages correctly.
  5. You'll see the chi-square results. The results will be clearly labeled as "chi-square", not "binomial", so you'll know for sure that you bypassed the problem.

This was fixed in 6.0b. 

Explore the Knowledgebase

Analyze, graph and present your scientific work easily with GraphPad Prism. No coding required.