What Are Confounding Variables?
A confounding variable usually appears in the cause-and-effect relationship and influences both independent and dependent variables. It is unmeasured but it can affect the results of research, so you should control it effectively if you want to obtain valid outcomes.
What Stands Behind a Confounding Variable?
A confounding variable or we can call it a confounder or a confounding factor, is an extraneous variable that meets two conditions:
- It correlates with the independent variable.
- It causally relates to the dependent variable.
Suppose we want to measure the effect of stress (IV) on gaining weight (DV). We need to ensure that confounding variables will not influence the measurements and results. Such variables can involve age, height, physical activity, gender, metabolism levels, types of stress, and individual reactions.
We start collecting data on stress and gaining weight. You think that higher levels of stress result in more food consumption. Does it necessarily mean that it results in gaining weight?
Here, individual reactions to stress make up a confounding variable. Stress makes some people eat more and consume more calories, while in others, it produces an entire loss of appetite and apathy leading to losing weight.
Why Is a Confounding Variable So Important?
Confounding variables can decrease the internal validity of the research. So, you should control them. If not, the results will not account for the actual relationships between the variables you are testing.
You may see a cause-and-effect relationship that does not exist because of the influence of a confounding variable. Suppose you have found out that more students who love pizza produce better academic results than those who don’t like it. Does this really mean that the love for pizza stimulates better academic performance?
Actually, not. It may be that students who love pizza feel more confident and satisfied with their student life, so they have more motivation to study better. You should consider other factors that influence their academic performance, such as levels of income, previous backgrounds, hobbies, etc., to understand the true causes of good academic performance, or you will arrive at conclusions that do not exist.
You may even identify a cause-and-effect relationship correctly, but when there is a confounding variable, you can underestimate or overestimate the true reasons for an independent variable to influence the dependent variable.
You may find that students who often miss classes do feel not healthy. However, you need to consider other behavioral patterns of such students and their reasons for missing classes - poor physical and mental well-being, bad habits, like drinking or taking drugs, disorganization, etc. - that lead to missing classes.
Ways to Reduce the Impact of a Confounding Variable
You can choose among a few methods to eliminate the influence of confounding variables on the results. These methods are applicable to different types of subjects under research - animals, people, plants, or substances. Every method has its pros and cons, so you need to consider them carefully.
Restriction
While using this method, you offer your treatment group only the subjects that have the same values influenced by confounding variables. These values do not differ, so they cannot correlate with the independent variable and affect the dependent variable.
You are planning to study whether the levels of income influence exercise intensity in office workers. You know that such factors as age, gender, and level of education can impact the ways people choose to exercise or not. So, you restrict your subject pool to 35-40-year-old men with Master’s degrees only.
The pros and cons of this method look as the following:
- ✔️ it is much easier to use than others
- ❌ it shortens the samples to a big extent
- ❌ you may not take into account other possible confounders
Matching
Here, you need to pick a comparison group that is pretty similar to the treatment group. Every participant from the comparison group needs to match a counterpart in the experimental group with the same possibility of confounding variables, though the independent variable values have to differ.
The possibility of the appearance of confounding variables that may cause variations reduces. You can consider the slightest differences in independent variables that lead to changes in dependent variables.
In the research on income influence on exercise intensity in office workers, you can match people with different ages, genders, and levels of education. It allows for extending the range of subjects. You will find a match for every 35-year-old woman with a Master’s degree and every 40-year-old man with a Bachelor’s degree and can do that for all the members of your sample.
The pros and cons of this method include the following:
- ✔️ you can involve more subjects in your research
- ❌ you may experience difficulties in finding the exact matches for all the potential confounders
- ❌ you may face other confounding variables that you cannot match
Statistical Control
You can turn confounding variables into control ones after you have collected all the data by using the regression model. You will be able to control the influence of confounders in this way. The effects a confounding variable may have on the dependent variable can be apparent in the regression outcomes, so you will be able to distinguish between the real status of the independent variable and its characteristics influenced by a confounder.
When you finished collecting data about the levels of income and exercise intensity in office workers from your participants, you can check the regression model and what control variables (confounders) you have included in it - age, gender, and levels of education, alongside the amounts of exercises every participant does per week. Each control variable relates to the independent variable - the amount of income. You can separate the correlation of the three confounders with it in every single case, and see the objective results due to this regression.
The advantages and disadvantages of this method are the following:
- ✔️ you can apply it easily
- ✔️ you can use it after the process of data collection is finished
- ❌ you can control only those confounding variables that you know of, while others may appear and influence the results without your consideration of them
Randomization
You can also randomize the independent variable values to diminish the effect of confounding variables. You can assign the participants to the groups randomly to ensure that all potential confounders are spread equally between these two groups and that their values and influences are also equal. This method works well for larger samples. These variables do not differ much between the groups, so they cannot correlate with the independent variable and influence the results of the study. Sometimes, choosing this method is the best way to ensure the research results’ validity by reducing the negative impact of confounders.
The group of subjects in your research about the correlation between income and exercise intensity in office workers is large. You divide them randomly between the experimental and control groups. It is a guarantee that such confounders as age, education levels, and gender, as well as other confounding factors you may not be aware of, are spread equally between the participants in both groups.
The pros and cons of this method can be highlighted in the following:
- ✔️ you will be able to consider all the possible confounders, even those you do not know about yet
- ✔️ you can sufficiently minimize the influence of any confounding variables
- ❌ picking out large groups is difficult to implement
- ❌ the process of many participant involvements should begin long before the start of data collection
- ❌ you need to check that only the participants in the treatment group but not all the participants of the research obtain the necessary treatment
Final Thoughts
When you know what to expect from confounders, you can prevent the negative influences of such variables on the outcomes of the experiment. In this way, you will ensure more internal and external validity of your study’s results, and can build up a solid basis for your further research.
We hope that the methods of reducing the influence of confounding variables on the results of the experiment will help you achieve more accurate results and consistent conclusions.