The Pareto platform is producing a sensible answer for the data that you are giving it. It is looking at your count data and telling you that of all the counts, test1 accounts for 93%. A pareto analysis expects the cummulative probability to sum to 100% - but that is not what you are asking for. Put another way, it expects all of the counts to be generated from the same sample of data e.g. number of counts by classification of defect for a batch of products. You data is more like a count of defects from 3 separate batches. What you want to do is to produce a graphical representation of the counts for the different groups. You could use the Chart platform to plot the percentages that you have calculated: use the "Data" statistic computed on your % column, with name as the category.
That's a good point! Sorting the data table by the % column doesn't change the display order of the bars. I don't know if there is another way??
So, plan B. Back to the pareto plot. There is a "weight" role in the pareto plot dialog. You can use this role to adjust for the fact that the different counts are derived from different size totals. Create a new column and call it something like weight. Assign the following formula
weight = 1 / total
(or weight = 100 / total ) depending on whether you want fractions or percentages.
Assign column 'name' to 'y,cause' Assign column 'weight' to 'weight' Assign column 'count' to 'freq'