Continuous data can be measured on a scale. Analysis of continuous data will paint a picture of the characteristics of your process measures and process outputs.
Overview: What is continuous data?
There are two categories of data:
- Discrete data, which is categorical (for example, pass or fail) or count data (number or proportion of people waiting in a queue).
- Continuous data is data that can be measured on an infinite scale, It can take any value between two numbers, no matter how small. The measure can be virtually any value on the scale.
Take length, for example. 100.2345 inches makes sense. Length is a continuous measure. However, the count of 9.5 people standing in a queue doesn’t make sense (half a person?). Count is a discrete measure. Measures of time, height, temperature, and thickness are all examples of continuous data.
2 benefits and 1 drawback of continuous data
There are a few important things to know about continuous data.
1. It provides information about the center, spread, and shape of the process measure sample
Continuous data can be summarized with descriptive statistics. You can calculate the average (center) and the standard deviation (spread). You could also calculate a measure of skew and kurtosis (shape). When you plot a histogram of a continuous data sample, a picture of the process measure emerges that tells you more than the statistics ever could.
2. It requires less data when used in graphic analysis and statistical tests
Continuous data is efficient — a little tells you a lot about the data. This is great news if your data is hard or expensive to collect.
3. It’s only as good as the measurement system that generates it
I call this a drawback, but measurement system analysis is really the price of entry for continuous data analysis. Most measurement equipment needs upkeep to provide data that is trustworthy. It’s your responsibility as a continuous data analyst to study and correct any issues with your measurement equipment prior to analyzing the data it provides.
Why is continuous data important to understand?
You must determine if the data generated by processes measures and/or process outputs is continuous in nature.
To choose the right statistics to describe the sample
As mentioned in the previous section, the descriptive statistics for continuous data include the average, standard deviation, skewness, and kurtosis.
discrete data may be summarized by counting occurrences of each category. You might also like to calculate the proportion (or percentage) of occurrences of a category in a sample.
To choose the right analysis tool
The tool you want to use in graphic or statistical analysis will specifically require either continuous data or discrete data.
If, for example, you accidentally use discrete data for a tool that requires continuous data, you can draw incorrect conclusions from the tool’s output. If you act on those incorrect conclusions, you may not get the results that you wanted, wasting both time and money.
An industry example of continuous data
Continuous data (temperature) from a curing oven was used to check if the curing oven could be used for a new product. The engineer needs the curing oven temperature to be centered around 200 degrees Fahrenheit and within 190 and 210 degrees Fahrenheit for the new product to be cured correctly. The engineer needs to:
- Verify the temperature gage with a Gage R&R Study. Assuming the gage passed…
- Take 30 temperature readings over the course of one day of curing the new product.
- Plot the data in a histogram.
The histogram of the 30 continuous temperature data points has a mean of 199.28.
What can this continuous temperature data tell us about the curing process?
- The center of the continuous data is close to what the engineer wants from the curing process. That’s good news!
- There are many continuous data values below 190 degrees Fahrenheit and above 210 degrees Fahrenheit. That is not good news.
- The process is not meeting the engineer’s requirements (specifications). Sometimes the curing oven is too cool, and sometimes it’s too hot.
Based on what they learned from the continuous data plot and statistics, the engineer decided to take action to reduce the variation of the curing oven temperature.
A mechanical check of the oven showed a thermostat was not functioning. It was replaced. The variation shrank back to acceptable levels, and the curing oven was again good to use with the new product.
3 best practices when thinking about continuous data
If you want to analyze data like an expert, keep these three things in mind.
1. Use Excel or a statistical and graphic computer program to analyze your continuous data
The days of plotting continuous data in histograms and calculating continuous data statistics by hand are long past. Find an analysis program that suits your needs and your budget.
Your company may allow you to download Minitab, JMP, or Excel. If not, you can search for free, open-source statistics software on the web. “R” software, for example, is free and used by many universities.
2. Assess your data for stability before you start analysis of continuous data
Before you use continuous data to represent your process measure or outcome, it’s important you know whether your process is in statistical control.
If your continuous data plot is not stable, you should do some process improvement work to move it toward stability.
Analysis of continuous data that is unstable only applies to that sample of continuous data. If the process is in statistical control, the analysis of the continuous data may also be applicable to the process samples from the near future.
3. Plot the data, plot the data, plot the data
A picture is worth a thousand words. Statistics support the graphs — not the other way around. Always begin continuous data analysis with descriptive statistics, a histogram or boxplot, and a control chart.
RELATED: DISCRETE VS. CONTINUOUS DATA
Frequently Asked Questions (FAQ) about continuous data
What graphs are good to use with continuous data?
There are many! Histograms, box plots, control charts, and scatter diagrams are some of the most popular.
Can continuous data be transformed into discrete data?
Yes! You can do this by setting upper and/or lower specification limits for the continuous data set. Any data point above the upper specification limit or below the lower specification limit can be called defective. Anything between the upper and lower specification limits can be called non-defective.
You now have a discrete data set called with two categories — defective and non-defective.
Some final thoughts on continuous data
What you learn when you analyze your continuous data may surprise even the most experienced process expert. As W. E. Deming used to say: “In God we trust. Everyone else, bring data.”
