dcsimg
 

Tally Hoe: Use Tally to Weed Out Those Pesky Typos in Data

The other day a colleague of mine mentioned a use for Minitab's Tally command that I didn't know about. I thought I would share it with you in the form of contrived anecdote. I hope you enjoy it.

Bob works at a manufacturing plant. You may remember Bob as the guy who was obsessed with gnomes. Well, his work with gnomes was so exemplary that he was promoted to work in the Garden Tool Division. Bob was sad to leave his gnomes behind, but the opportunity was too good to pass up, so Bob traded in his gnomes for rakes, shovels, and hoes.

When Bob got to the tool plant, he found that something was amiss with the hoe production. Too many hoes were being scrapped, so he decided to investigate. Now when a hoe is scrapped, you are supposed to record the reason as one of the following 6 defects:
 

Defect Description
Bent blade The flat metal part of the hoe isn't so flat.
Bent stem The metal part that connects the blade to the handle is bent.
Knotty handle The handle stock has knots in it that will eventually lead to breakage of the handle, possibly at an inopportune time.
Loose hilt The metal collar that holds the stem firmly on to the handle is not tight enough.
Split handle The handle has started to split, to cleave, to begin the process of becoming 2 separate entities, unconnected one from the other.
Splintering handle The handle isn't finished properly, such that bits of wood are protruding and may find their way into a customer's skin, perhaps at an inopportune time. (The label says "Always wear gloves and safety goggles when operating this equipment," but no one ever does.)

  
So imagine Bob's surprise when he created a Pareto chart of the defects and found this:
 
Pareto with "other" category
  
In addition to the expected 6 defect categories (with "Loose hilt" leading the pack), there was an unexpected category for "Other." Bob wasn't excited about wading through hundreds of rows of data looking for the "other" defects. Fortunately, he knew a better way.

Bob chose Stat > Tables > Tally from the Minitab menu, entered the column with the defects data, clicked OK, and was immediately confronted with the problem that had caused the mystery defects—a loose kilt.

Loose kilt?Actually, there were a few problems. It is possible that the same conditions that led to the "Loose kilt" caused that person to "Lose" a kilt. It may have been a major distraction on the production floor.

And it seems that others were affected by the distraction as well, as evidenced by a "Snotty handle" and also the "Spit handle." Find it here

Bob decided he needed to clean things up. He knew that Minitab had a handy Find feature, so he clicked on the worksheet to make it active and then chose Editor > Find from the menu.

Bob typed in "kilt" and quickly found the problem cells. He cleaned up the "snot" and "spit" in the same way. And when he recreated his Pareto chart, it was as he expected, with no mysterious kilts, spit, or snot.


Better Pareto
  
And with that, Bob was happy. For a time ....
  

Comments

blog comments powered by Disqus