Attribute data vs variable data. If you had a choice, I’m sure you’d pick variable data!
I often find Six Sigma practitioners shy away from projects that try to reduce non-conformances. They’d rather work on projects that have a well-defined output variable instead of an attribute. I don’t blame them; I’d rather have a precise output variable as well. It’s so much easier to use a gauge, measure a feature of interest, and compare it to a specification. With variable data, I can quickly sample production units, collect my data, plot the data, and interpret the plot in a short time.
Attribute Data vs Variable Data
Attribute data is a different ball game. I need to collect many samples, count the defects in that sample, and then compute a defect rate. And for all this effort I get 1 data point. This single data point consumes a lot of energy and time!
Defect rates are often computed by shift, day, week or month. Let’s assume we have the defect rate by day. Let’s also assume we have a remedy to reduce the daily defect rate. If we were to apply that remedy, we would need at least 7 to 9 data or 7 to 9 days to confirm the effect of that remedy. If we had variable data, we could immediately know if a remedy had merit. First, we would find the sample size needed to detect a difference. Second, we would collect that many samples and measure the feature of interest. We would then compare the pre and post statistics and rend a decision. In this case, we would know the answer that day.
Variable Data Takes Less Energy
The point I am trying to make is variable data gets you the information you need sooner. You get 1 data point for each unit produced. The ratio of data to units is 1 to 1.
Attribute Data Takes More Energy
Attribute data takes many samples to compute a defect rate. Here the ratio of data to units is 1 to many units. And once you need many units to compute a single value it eats up a lot of energy and time. And for this reason the improvement project will take more energy and time to complete.
Attribute Data vs Variable Data
So, what is a Six Sigma practitioner to do? He or she is working against the clock. Their manager wanted the project completed yesterday and there is no way the project is going to get done by tomorrow!
So here is my secret. Are you ready for it? Turn your attribute into a variable. What? Can you do that? The answer is YES when you use a technique called PSYCHOMETRIC SCALING. To better understand, let me share the following story.
I discovered Psychometric Scaling while having an eye exam many years ago. For those of you that wear eye glasses I’m sure you’ll appreciate this.
My Optometrist Uses Psychometric Scaling Everyday
While sitting in the exam chair, I had a large contraption placed on my face. This contraption has various lenses the Optometrist uses to figure out my prescription. The Optometrist presents the first lens followed by the second lens. During this time he asks, “better or worse?” He continues to present pairwise lens and asks the same question, “better or worse?” Once the examination is over, the Eye Doctor writes a numeric value for corrective eye glasses. With prescription in hand I asked, “How can you go from a pairwise comparison of lens and derive a number to two decimals places?” That question led to the study of Psychometric Scaling and a method called Paired Comparisons.
I soon discovered several techniques – Steven’s Power Law, Successive Categories and Magnitude Estimation. My favorite is the method of Paired Comparisons.
To see the Method of Paired Comparisons applied in practice, you can ask for a copy of a paper I wrote several years ago. It used the technique in the analysis of automotive seating comfort. In the article I show how I took the subjective evaluation of people and turned an opinion into variable data.
Today I don’t fret over the choice of Attribute data vs Variable data. When faced with an Attribute I use the right psychometric technique that helps me turn a subjective opinion into an objective number.
Would you like to learn more about Psychometric Scaling? Then let me know if you’d like to see an example! If I get 15 or more requests in the comments section below then I’ll write another post.