Concomitant variables in statistical analysis are observed during the study but are not the focus of the research.
Concomitant variables could affect the variables being studied and skew or bias the data. Researchers must often correct variances in these secondary variables to produce reliable results. They are sometimes also called covariates.
Pros and Cons
Good use of covariates is more complicated than tracking all the variables that might affect a process. Studies must be conducted in the real world, and sometimes data may not be worth the cost of acquiring it. Good experimental design requires significant balancing between accuracy and cost.
Pro: Defining and measuring a large range of concomitant variables helps bring precision to a study. In Lean Six Sigma management, where you’re attempting to reduce defects to an extremely low level, precision is a must.
Pro: Listing the variables that affect a process helps the researcher better understand what is occurring in the study.
Con: Each variable you track in a study increases its cost and complexity. Introducing more variables to a survey also introduces more sources of possible error. While noting everything that might affect the process you’re researching may sound great in theory, it can be more trouble than it’s worth.
Con: Introducing a sheaf of concomitant variables in the interest of precision may distract analysts from the key factors that drive the process or make the study so hard to conduct that its accuracy suffers.
Why Are Concomitant Variables Important to Understand?
The bottom line is that ignoring concomitant variables can render an expensive study completely worthless. Conversely, tracking too many variables can make research cumbersome and difficult to conduct.
Making sure to get them right is an important part of sound experimental design. Paying attention to the wrong variables leads to unreliable results or unsound analysis. Further, good experimental design can eliminate many of the effects caused by covariates.
Your list of concomitant variables is good starting point to look for problems if the results of your study are unsatisfactory or show high variance between trials.
Industry Example of a Concomitant Variable
Suppose analysts want to know if a fertilizer increases crop yeilds. However, there could be regional variations in sunlight, rain frequency, and differences in the volume of water used by farmers. These variables could also have significant impact on crop yeilds.
Four Best Practices When Thinking About Concomitant Variables
The following tips will help you use covariates in your studies in an efficient manner:
1. Domain expertise in the process under study helps identify concomitant variables.
2. Pay close attention to the relative effect a variable has on the process in question.
3. Also look at the number of covariates at play.
4. It’s more difficult to eliminate a concomitant variable from observational studies than experimental studies.
With observational research, analysts often must note the covariates and can do little to eliminate or limit them. However, experimential research can use (for example) randomized samples to minimize the impact of covariates.
Frequently Asked Questions About Covariates
Using concomitant variables in Lean Six Sigma management studies frequently lead to numerous questions:
How do you account for the effects of concomitant variables?
Why is it more difficult to eliminate or limit concomitant variables in observational studies?
How does randomized samples limit the impact of a concomitant variable?
Balancing Accuracy Against Study Cost and Efficienty Is a Judgement Call
No study is going to be perfect. You will always make tradeoffs between the cost of the study and its accuracy. Insist on too much precision, and the analysis can be too expensive or too difficult to execute to provide meaningful data. However, failing to identify and adjust for covariates can make a study’s results unreliable or even meaningless.