Confounding variables are common in research and can affect the outcome of your study. This is because the external influence from the confounding variable or third factor can ruin your research outcome and produce useless results by suggesting a non-existent connection between variables.

In order to control confounding variables in research, it is important to know how to clearly identify these third factors, and know how they influence your research outcome. Understanding and controlling confounding variables will help you achieve more accurate results in your research.

A confounding variable, in simple terms, refers to a variable that is not accounted for in an experiment. It acts as an external influence that can swiftly change the effect of both dependent and independent research variables; often producing results that differ extremely from what is the case.

In correlational research, confounding variables can affect the perceived relationship between the 2 variables under consideration; whether positive, negative or zero. A confounding variable may also be defined as a factor that a researcher was unable to control or remove, and it can distort the validity of research work.

There are several research methods used to identify confounding variables. The most common method is to observe the extent to which removing a factor in research causes the coefficient of other independent variables in the research to change.

In this sense, the researcher observes and measures the estimated level of association between independent and dependent variables, both before and after making adjustments. If the difference between the 2 measuring parameters is more than 10%, then a confounding variable is present.

Another method of identifying a confounding variable is to determine if the variable can be linked with both the exposure of interest and the outcome of interest in research. If there is a meaningful and measurable connection between the variable and the risk factor and, between the variable and the outcome, then such variable is confounding.

There are several hypothetical and formal testing methods for identifying confounding variables. The baseline model, biological model, and binary logistics and multivariate logistic regression models are some of the common research hypothetical methods use to identify confounding variables.

Confounding variables can also be identified using various tests of co-linearity such as measuring variance inflation factors. You can calculate the variance inflation factor for all variables in your research to see if this indicator is high for any of the variables; hence indicating a confounder.

Confounding variables are often associated with both the risk factor of interest and its outcome. They are usually distributed unequally among the independent and dependent variables in research and, confounding variables do not lie between interest and outcome in research.

A confounding variable can function in 3 possible ways in the research: as a risk factor, as a preventive factor or as a surrogate or marker variable. The common formulae for calculating the extent of confounding in research are:

- Degree of Confounding = (RRcrude – RRadjusted)/RRcrude
- Degree of Confounding = (RRcrude – RRadjusted)/RRadjusted

Confounding variables cut across several fields of study especially Statistics, Research Methodology, and Psychology. In all of these fields, these third factors retain their primary characteristics of extremely influencing the research outcomes of dependent and independent variables from outside the controlled environment.

**A Mother’s Career**

Research is carried out to determine the degree of correlation between formula feeding for babies and intelligence in infants. It would appear logical that children who are formula-fed may be less intelligent because they do not get the nutrients, vitamins, and minerals contained in breast milk.

However, the facts may be that formula-fed kids can be even more intelligent than children who are breastfed. Infant feeding formula actually contains nutrients, vitamins, and minerals that can help to boost a child’s intelligence and help protect the child from infections.

A confounding variable in this instance can be the mother’s career, that is if the mother is a housewife or working-class mother. When this 3rd factor is considered, you would find out that working-class mothers are more likely to choose formula-feeding because their jobs may not allow them to always breastfeed their infants.

**Weather**

Research is carried out to determine the extent of correlation between money and the sale of ice-cream. Logic may suggest that there is a positive correlation between these 2 variables; that is, people buy more ice-cream when they have more money.

A confounding variable in this research can be the weather so that it is distinctly possible that the weather is the correlative causative factor. Thus, when the weather is cold, people work less and have less money to buy ice-cream and, when the weather is hot, people work more and have more money for ice-cream.

In this example, the variable causing the relation between money and ice-cream is the weather.

**Slanted Plank**

Statistically, a confounding variable may come into play in the case of the relationship between the force applied to throw a ball and the distance that the ball travels. Logically, it is assumed that the more force exerted on a ball; the farther it would travel.

The confounding variable, however, would be what direction the ball travels on a slanted plank. If the ball is traveling upwards, it may travel slower regardless of the force and, if it is traveling downwards, it would travel faster with little force exertion.

**Eating Habits**

A confounding variable may also be responsible for the correlation between exercise and weight loss. The natural logic may be that the more you exercise, the more likely you are to lose weight, but a confounding variable in this research may be eating habits.

This means that the more people eat, the more weight they gain and vice versa.

Read: Dependent vs Independent Variables: 11 Key Differences

**Sleep**

Psychologically, a confounding variable can influence the connection or relationship between caffeine and concentration. You notice that the more caffeine you take, the better concentrated you are in class; here, concentration is dependent on the level of caffeine which is the independent variable.

The confounding variable, in this case, could be sleep; that is, you may have been getting better sleep leading to better concentration levels, irrespective of the level of caffeine consumption. The confounding variable, in this case, has nothing to do with the research independent variable; that is, caffeine consumption.

To properly understand the effect of confounding variables on dependent and independent variables in research, it is necessary to understand what dependent and independent variables are. This will help you to clearly contextualize both of these research factors.

An independent variable is a lead factor that triggers a change in the other factors in the research environment. In the same vein, the dependent variable is the factor that is acted upon in research, and it results from the influence of an independent variable.

A confounding variable can affect the correlational relationship between independent and dependent variables; often resulting in false correlational relationships as it may suggest a positive correlation when there is none. It can also trigger an extreme change in a dependent variable and consequently, the research outcome.

In terms of the independent variable, a confounding variable or confounder can affect this factor by removing it from the correlational relationship and research process, entirely. This happens when the research outcome results from a change triggered by the confounder rather than the independent research variable

Confounding bias is one of the effects of having confounding variables or third factors in your research. It is the result of a distortion in the degree of association between an exposure and the outcome if exposure in research.

A confounding bias may be negative or positive in nature. In this sense, a negative bias underestimates the outcomes of research while a positive confounding bias overestimates the outcome of research; causing an accelerated distortion of the outcome.

Confounding bias occurs when a research data set is corrupted through poor gathering techniques such that the entire research process in itself is set up without enough controls. This allows for confounding variables to arise and affect the research outcome.

It is important to limit or control the effect of confounding variables or confounders in the research process. Usually, a researcher can only control or ultimately avoid confounding variables in research when he or she can identify and measure the possible third factors in the research environment.

There are 5 common strategies for reducing or avoiding confounding variables. These are:

**Randomization**

The randomization method involves distributing confounders across your research data sporadically. It is used in machine learning to randomly assign variables to a control group in research and it helps to prevent any cases of selection bias in research work.

Randomization is usually adopted in experimental research to enable the researcher to control these variables. It redirects the experiment from looking at an individual case to a collection of observations, where statistical tools are used to interpret the findings.

A random sample is a type of sample in which every member of the sampling group has an equal chance to be sampled. It is important to note that a perfectly random sample of observations is difficult to collect and so, the researcher has to work to achieve randomization as closely as possible.

**Restriction**

This method limits the research to the study of research variables with control for confounding variables, and, if it is not done carefully, it can lead to confounding bias. It involves restricting the research data by introducing control variables to limit confounding variables.

**Matching**

The matching method distributes the confounding variables across the research data, evenly; by using a controlled research process like before and after experiments. It involves making observations in pairs; one for each value of the independent variable that is similar to a possible confounding variable.

A common method of matching is the case-control study that involves matching variables of similar characteristics with the same set of controls. A case-control study may have 2 or more controls for each case, as this gives more statistical accuracy in your research process.

**Stratification**

Stratification is a method of checking the activities of confounders by distributing these factors equally at each level of the research data analysis. It involves dividing the data sample into smaller groups and examining the relationship between the dependent and independent variables in each group.

**Multivariate Analysis**

This method is entirely dependent on the researcher’s ability to identify and measure all the third factors in the research.

Other tips include counterbalancing by introducing different research analysis parameters, where half of the group is examined under condition 1 and the other half is examined under condition 2. You can use the “within-subject method” to test the subject each time as in-between periods can trigger confounding variables.

Although somewhat similar, there is a fundamental difference between a confounding variable and an extraneous variable. It is important for every researcher to be able to clearly recognize this difference in order to accurately identify the variable acting in a research outcome.

An extraneous variable is a type of variable that can trigger an association or correlation between 2 research variables that have no causal relationship. If the relationship between the 2 variables; A and B, is caused solely by a 3rd variable; C, then such a relationship is spurious and variable C is an extraneous variable.

A confounding variable, on the other hand, affects 2 variables that are not spuriously related, that is, not solely related by the 3rd factor. In this case, the relationship between variable A and B is already causal, that is, A causes B.

When the causal relationship between variable A and variable B is also influenced by a third variable C, variable C is a confounding variable. Thus, the association between A and B may exaggerate the causal effect of A on B because the association is inflated by the effect of C on both A and B.

Confounding variables can result in 2 extreme research problems which are increased variance and research bias. Each of these effects will be fully considered below, and they can largely tilt your research result to be overestimated or underestimated in the end.

**Increased Variance**

Increased variance refers to an escalation in the number of possible causative and independent variables in research. This is common with research that does not have any control variables such that the changes in the dependent variable can be triggered by other variables.

For example, your research reveals that increased weight gain results from a lack of exercise. However, because there are no control variables, you cannot trust your research outcome because there are a number of factors that can affect the dependent variable.

For example, one of the confounding variables, in this case, can be genes or genetic factors. Another confounding variable can be an Individual’s eating habits so there are too many possible causative factors that end up distorting the results.

**Confounding Bias**

A confounding bias refers to the chances of a statistical parameter to overestimate or underestimate a research parameter. A survey design that has clear occurrences of confounding bias could lead to high survey dropout rates and survey response bias which affects the research outcome.

A confounding bias may be positive or negative in nature and can ruin the internal validity of an experiment. A positive confounding bias occurs when the observed association is biased away from the null such that it overestimates the effect.

On the other hand, a negative confounding bias occurs when the observed association is biased toward the null in such a way that it underestimates the effect. Negative confounding bias can lead to a false rejection of a null hypothesis.

**Wrong Research Outcomes**

A confounding variable can alter the outcomes in research. As an external variable, the third factor can change the effect of both dependent and independent variables in research; thereby influencing the outcome of correlational or experimental research.

Since a confounding variable is a 3rd factor that is not accounted for in a research process, it can affect an experiment by producing inaccurate research results. For example, it can suggest a false correlational relationship between dependent and independent variables.

Although third factors are usually considered as invalid variables in a research process, they can change the course of a research by reflecting false correlational relationship between variables. Hence, it is necessary to always control your research environment in order to reduce the effects of confounding variables.

In this article, we have highlighted 5 simple and common control methods for confounding variables including randomization, matching, stratification and restriction. These strategies would help you to better manage your research outcomes better by limiting the effects of third factors.

You may also like:

In this article we’ll look at the different types and characteristics of extrapolation, plus how it contrasts to interpolation.

This article will discuss the two different types of errors in hypothesis testing and how you can prevent them from occurring in your research

In this article, we are going to discuss extraneous variables and how they impact research.

In this post, we will discuss extensively what acceptance sampling is and when it is applied.