The PARETO Procedure

Pareto charts shown in textbooks usually scale the cumulative percentage curve so that it is anchored at the top right corner of the leftmost bar. The upper end of the frequency axis is then extended to accommodate the curve. For an illustration, see Figure 15.1. By default, the PARETO procedure uses the top right corner as the anchor position on a vertical chart and the bottom right corner of the topmost bar as the anchor position on a horizontal chart. You can override the default by specifying the ANCHOR= option.

This method of scaling is not feasible if the number of categories is very large and if the Pareto distribution is uniform. In this case, the bars are excessively compressed relative to the curve. Conversely, this method excessively compresses the curve relative to the bars when you use a count scale for the frequency axis in a comparative Pareto chart and the tallest bar does not occur in the key cell. In either situation, PROC PARETO overrides the textbook scaling method and balances the scales of the bars and the curve.

You can use the AXISFACTOR= option to specify the extent to which the frequency axis should be extended. Alternatively, you can extend the frequency axis by using the FREQAXIS= option to specify the tick mark values for the axis.

Another scaling anomaly is illustrated by the comparative Pareto chart in Output 15.1.4. There, the cumulative percentage curve in the bottom chart is not anchored because a uniform count scale is combined with different sample sizes in the two cells.