A run chart is a graphical display of data plotted in some type of order. It uses the median as the midline. It may also have a goal line. It is a visual tool that encourages users to ask good questions based on the data they are viewing.
① Simple Example
Run charts make variation visible. Although this project has documented recent improvement approaching the goal (95%), the chart should stimulate comments or questions about the April data point (68).
② Leading to Questions
Run chart with goal line and tests of change annotated. Protocol version ① with no response and protocol version ② resulting in a shift (8 data points above midline) should have annotations on the chart at the appropriate point in the timeline.
③ Non-random Variation
Studying a run chart can often reduce the risk of over-interpretation or tampering with the process. Does the chart show evidence of improvement that is obvious to all viewers? Use a set of rules to detect nonrandom change in the data.
④ No Evidence
24 data points not on median. This chart has 23 crossing + 1 = 24 runs (expect 8 ~ 18 runs). This is too many runs and is a signal of non-randomness. This pattern most often represents the need for stratification.
⑤ Too Many Runs
11 data points (1 on median). This chart has 1 crossing + 1 = 2 runs (expect 3 ~ 10 runs). This is too few runs and is a signal of non-randomness, and is a signal of improvement if the change headed in the desired direction.
⑥ Too Few Runs
The median should be created using 10 or more data points. If the median exhibits no signals of non-randomness, extend it into the future. This allows better detection of signals of improvement.
⑦ Extended Median
When analyzing a run chart with two medians, the rules must be applied separately to the data surrounding each median. Baseline data to ① for the first median; improvement strategies from ① to ②, then the second median calculated from ② to ③
⑧ Two Medians
When more than half of the data points are on the median, the probability-based rules can not be used; the mean should be used. However the shift and run rules should not be applied.
⑨ Too Many Points on Median
If more than half the data are at the extreme values on the scale, detecting a non-random signal using the median cannot be applied. The median will be the extreme value itself (0 or 100%). Use the mean or change to time-between interval charts.
⑩ Too Many Points at Extreme Value
If more than half the data are at the extreme values on the scale, detecting a non-random signal using the median cannot be applied. The median will be the extreme value itself (0 or 100%). Use the mean or change to time-between interval charts.
⑪ Too Many Zeros
Place a trend line only if that chart provides a signal of change and the data appears to move in a consistent downward or upward dirction on the run chart. Can be used to estimate when goal will be reached.
⑫ Trend Line
