Extraneous and potentially confounding variables are a type of experimental research.
There are numerous ways of minimising extraneous and potentially confounding variables:
The method of sampling utilised can be used to minimise extraneous variables. A sampling method is often chosen depending on what the extraneous variables which are being minimised are. Sampling procedures include convenience sampling, random sampling, stratified sampling, and random stratified sampling.
Convenience sampling involves selection merely by what is quick and easy. It makes no attempt for the sample to be representative of the population, and is often biased. Whilst convenience sampling is easy (an experimenter may take, for example, one whole class out of a school of 1000 students), results cannot be generalised. Convenience sampling comes with many possible extraneous and confounding variables.
In random sampling, each member of a population has an equal chance of being part of the sample. Extraneous variables are likely to be minimised, assuming that the sample is large enough.
When a particular characteristic is important to a study, stratified sampling may be used. Stratified sampling may eliminate extraneous variables regarding a particular characteristic. Using this sampling method, the proportion of participants with a particular characteristic (red hair, for example) is the same as the proportion of participants with that same characteristic in the wider population.
Random stratified sampling is a more elaborate method of stratified sampling. Each member of a population with a particular characteristic (red hair, for example) has an equal chance of being selected to be part of the sample. As a result, the sample will consist of proportionate sizes of participants with particular characteristics to the wider population. Random stratified sampling is the most accurate sampling method, which means that it is likely to best minimise extraneous variables, but it is also the most time-consuming.
Case studies involve observation of an individual or individuals over a period of time. The findings cannot be generalised as each case is specific.
Observation involves natural observation (watching the subject behave in their natural environment), which is realistic but uncontrolled, and controlled observation (watching the subject behave in an ‘artificial’ environment), which can be manipulated but is less realistic.
Interviews include structured interviews, where there are pre-determined questions and fixed responses, and clinical interviews, where there is more flexibility.
Surveys are easy to compare, and allow for data to be quantified, but may produce limited or biased responses.
Psychological tests (an intelligence quotient test, for example) are easy to replicate and compare, but are subjective in validity.
Cross-sectional research designs can be utilised to minimise the potential confounding variable of time, as all data is collected at once: data of participants of all ages is collected at the same time and then compared. Whilst cross-sectional designs are comparatively easy and quick, a lot of participants are required to achieve accuracy.
Conversely, longitudinal research designs can be utilised to minimise the potential confounding variable of age, as the same participants are studies over a long period of time. Although longitudinal designs help with analysis of development over time, it is much more time-consuming, and participants are more likely to withdraw from the studies.
Counterbalancing is a common method used to minimise extraneous variables between the experimental and control groups. In counterbalancing, half of the participants are exposed to the experimental condition and then the control condition, whilst the other half of the participants are exposed to the control condition and then the experimental condition
In a single blind procedure, either the participants or the experimenters are unaware which group is exposed to the experimental condition, and which group is exposed to the control condition.
In a double blind procedure, both the participants and the experimenters are unaware which group is exposed to the experimental condition, and which group is exposed to the control condition.
In an independent groups research design, there are two groups: the control group and the experimental group (exposed to the manipulated independent variable). Participants are allocating to the groups randomly. Independent groups designs are often vulnerable to participants effects.
In a matched-participants research design, all participants are tested in regard to a potential confounding variable before they are allocated to the control or experimental group. Once all participants have been tested, they are ‘matched’ into pairs. For example, if the potential confounding variable is height, the tallest and the second tallest participants would be paired, the third and the fourth, the fifth and the sixth, and so on. One participant from each pair is then allocated to the experimental group, whilst the other is allocated to the control group. Although this research design takes a lot of time (due to the added testing before groups are decided), the potential confounding variable should be eliminated.
In a repeated measures research design, all participants are exposed to both the experimental condition and the control condition, meaning that the impact of the independent variable can easily be analysed. Whilst this design eliminated participant effects (as the same participants are used), order effects are likely. To minimise this effect, counterbalancing can be used.
Want to suggest an edit? Have some questions? General comments? Let us know how we can make this resource more useful to you.