Choosing the Right Control Chart

Calls received by our tech support line follow the 80/20 rule. 80% of the calls have to do with one question: Which control chart do I use?

Finally, I had a BFO (brilliant flash of the obvious). QI Macros already analyze your data to figure out how best to graph your data, why not let them automatically figure out which control to use too? Guess what, it works. The Control Chart Wizard is now available in the most recent version of QI Macros.

If you are still choosing your control chart the old fashioned way, here are some tips.

Choosing a Control Chart

With the recently added XMedianR chart, there are now more than nine control charts in QI Macros. How do you know which one to use?

When I'm working with data in Excel, I follow a simple strategy for selecting the right chart based on the format of the data itself. There are three formats I look for:

  1. A single row/column.
  2. Two rows/columns with a numerator and a denominator.
  3. Two or more rows/columns containing multiple observations from each sample.

They look like this:
c control chart data p control chart data XbarR control chart data

Single Row/Column

If you only have a single row/column of data, there's only three charts you can use:

  • c chart (attribute or counted data) It's always an integer (e.g., 1,2,3,4,5).
  • XmR chart (variable or measured data) It usually has decimal places (e.g., 33.75).
  • XmR Trend chart for variable data that increases (e.g., rising costs due to inflation)

So which one should you choose? If you're counting indivisible things like defects, people, cars, or injuries, then choose the c chart. If you're measuring things like time, length, weight, or volume, choose the XmR chart. The data looks like this:

c Chart
XmR chart

XmR Trend

defects, people, cars, injuries
time, money, length, weight, or volume
Increasing or decreasing time, money, length, weight, or volume
c control chart data
XmR control chart data
XmR trend control chart data

Look for these patterns in the data and then select the chart.

Two Rows/Columns

If the data has a numerator and a denominator that varies (e.g., defects/batch, errors/transactions), then you will want to use the:

  • p chart (one defect maximum per piece)
  • u chart (one or more defects per piece)

How can you tell which one to use? I ask myself: "Can this widget have more than one defect?" If yes, use the u chart, otherwise use the p chart.

p Chart
u Chart
Defective items per batch
Defects per item
p control chart data
u control chart data

Sometimes, as in this example, you can have more defects than samples. This is another clue. Again, look for these patterns in the data and then select the chart.

Two or More Rows/Columns of Variable Data

Service industries don't use these charts very often. They are mainly used in manufacturing. If you have two or more rows or columns of variable data (time, weight, length, width, diameter, or volume) then you can choose one of three charts:

  • XbarR (Average and Range, 2-10 rows/columns per sample)
  • XMedianR (Median and Range, 2-10 rows/columns per sample)
  • XbarS (Average and Standard Deviation, 5-50 rows/columns per sample)

Your data should look like this:
XbarR control chart data

You can run the XbarR, XMedianR or XbarS on this data. Xbar uses the average as the measure of central tendency. The XMedianR uses the median. If you have more than five samples per period, then the XbarS will probably be the most robust chart for your needs. You can also use the XbarS if your data has a varying number of samples per period:
Xbar R control chart data with varying sample size

Again, look for these patterns in your data and then select the chart.

The np Chart

There's one chart I've left to last because I rarely find situations where it applies. The np chart is like the p chart except that the sample sizes are constant. The data looks like this:

np control chart data

Again, look for these patterns in your data and then select the chart.

"Group by Row or Column" Prompt

Foundationally, our software assumes you have more data points than samples. If you have more rows than columns, QI Macros will default to group by columns, and typically only asks “by row or column” if you have more columns than rows.

And remember that when using QI Macros, only select one row and one column of headings or labels.

Summary

So, just recognizing patterns in your data can make it easier to pick the right control chart.

Rows/Columns

Attribute
Variable
1
2

 

2 or more
 

If you learn to look for these patterns in your data, it will make it easier to choose the right control chart. And it's so easy to draw these charts with QI Macros, that you can draw them and throw them away if they aren't quite right. For further help, download our SPC Quick Reference Card or consider our 1 hour SPC Simplified Training video on DVD.

Please note: If you are asked by QI Macros for a "Subgroup Size," a subgroup is a group of measurements produced under the same set of conditions.

You can find all of this data in c:\qimacros\testdata or Documents\QI Macros Test Data.

Create these charts and diagrams in just seconds using QI Macros for Excel...
Try It Now!