The fraction of total analysis divides each value by its column or row total, or by the grand total.

 

This analysis is most often used for parts-of-whole data or for contingency tables, but it can be used for column data and for XY or Grouped data tables, so long as they have no subcolumns.

 

The calculation of confidence intervals only makes sense when each value entered into the table is an integer that represents the actual number of objects or events counted. In this situation, computing a confidence interval based on the binomial distribution makes sense. Don't check the option to compute confidence intervals if the data are measured values, or percentages, or normalized.

 

With the first three choices, the results are presented on a results table with green grid lines, so those results can be plotted. With the last choice (presenting fraction of all three totals), the results are on a table with red grid lines so cannot be plotted. With this last option, the option to report confidence intervals is not available.

 

Prism can compute the confidence intervals of proportions using three different algorithms. We recommend the Wilson/Brown method. Prior versions of Prism used the Clopper-Pearson method.