In the world, every corner has a population that has varying experiences from each other. Most of them cannot be generalised. And if selection bias leads to errors while forming conclusions, this article explains what it is and the disadvantages of its existence in research in a detailed manner.
What is it?
Selection bias is an error that occurs when the participants in a research study do not represent the entirety of the population. The selection isn’t random, and here, the researcher cannot understand the variables in opinions, and hence can result in misleading conclusions.
Examples:
- If a person decides to see the effects of a specific over-the-counter (OTC) medicine. But they only choose younger adults (above 20 and below 40), not the older population (above 70). This cannot tell that person about the possible effects on the latter.
- If a researcher focuses on students with higher marks, they will know that the subject is quite easy to understand. But they should focus on the entire class to understand their objective properly.
Types of bias:
- Sampling bias: This happens when the method for choosing participants/respondents includes only specific members of the population, but not the others.
- Self-Selection bias: Also known as Voluntary Response bias, it occurs when people choose to participate, but their response differs largely from those who don’t.
- Non-response bias: It occurs when people who participate do not do it actively or in a way that leads to productive findings. For example, people with low income levels may feel awkward while answering surveys related to their financial situation.
- Exclusion bias: It happens when certain groups of the population are excluded on purpose.
- Attrition bias: This bias occurs when people leave long-term research and surveys for any reason. And the ones who stay have different opinions compared to the ones who left.
- Time interval bias: It occurs when the parts of the population are selected during a period that can result in misleading conclusions. For example, if a person studies the stock market during an upswing, which may seem like growth, but isn’t.
- Survivor’s bias: This form of bias occurs when only successful cases and results are considered, but the opinions of the ones who were dropped or left are not.
How is selection bias formed?
In some cases, this mistake is either intentional, unintentional, or an oversight. Researchers may not even know that they are being biased until people point it out directly. Here are a few reasons for selection bias to occur:
- Lack of knowledge: If proper research about similar situations arent done, researchers may not know of previous biases. Hence, a proper understanding should be acquired.
- Accidental exclusions: Some people who would’ve participated if they knew about the survey. Hence, they might be excluded. Or if the survey was done online, only the ones who were informed could participate.
- Volunteers: People volunteering for a survey or research are a helpful cause, but it may attract those with specific characteristics that may or may not help research.
Implications:
Selection bias can undermine the whole picture, as it only takes a certain piece from the puzzle. In healthcare, if the clinics consider only the ones who attend their clinics, they won’t be able to gather the opinions of the ones who don’t or attend a different clinic. In education, the management cannot consider the top-achieving students as sole representatives, as other students may not perform well academically but possess sharp minds and problem-solving abilities in other fields. And in business, if the companies consider only their popular products, they may undermine the performance of the comparatively less sold ones, which can also be liked by many. Hence, selecting a specific part of the population to represent the opinion of the entirety can lead to wrong results.
Case studies:
- COVID-19: During the pandemic in 2020, only the symptomatic patients were considered for study and close observations. But the ones who were asymptomatic were not, as if the latter did not carry the virus at all. This undermined the effect and severity of the situation.
- Fitness stories: Particularly on social media, fitness groups advertise about the people who have succeeded in their programmes, but do not show the ones who didn’t, which can be demotivating, as success could be defined differently by the organisation and the individual.
Impact of Selection Bias:
- Research outcomes: If research is done with the influence of selection bias, it can overlook many important aspects of the objective. It won’t help either the researcher or the people to understand the entirety of the situation, and hence, end up spreading insufficient information.
- Reduced Validity: As the results focus on only a particular portion of a situation and are not well-rounded, they can be less credible. like only telling one side of the story, which can be misleading.
- Ethical concerns: If certain groups of the population are not involved, it can raise ethical concerns. It undermines inclusivity and human concern.
How to detect selection bias:
Though it is a bit complex, here is how one can detect the influence of selection bias in research:
- Compare the selected population to the population: Check if the sample population is an amalgamation of every part of the wider population.
- Check your selection process: See how people were chosen in your selection process and on what basis.
- Let people know about your research: Let your friends or a group of individuals whom you trust know about your findings, and let them analyse them.
How to correct selection bias:
One must always strive to decrease their bias about something, especially if it is intended to be widely spread knowledge. Here are some methods to reduce selection bias:
- Employ random sampling techniques: Researchers should avoid non-random sampling methods and must select the most random approach. Therefore, they should initially choose individuals without questioning their background, and after gathering them, they should fully explain their purpose before proceeding to ask questions.
- Stratified or Systematic sampling: Stratified sampling is where people are further divided into subgroups or ‘strata’, and Systematic sampling is where people are divided based on specific characteristics like height and age. This helps researchers to know their people beforehand and hence, pick a more diverse group that comes from all parts of the population.
- Pair similar people: Also, for better comparative analysis, pair up people from the same localities or age. This helps researchers to find various opinions from similar individuals.
- Select the misrepresented: A researcher should choose people who are usually ignored for a larger study, thereby decreasing bias but also obtaining new perspectives and opinions.
To conclude, selection bias can be a dangerous influence as it can cover up diverse opinions that can change a perspective on something. Incidental or accidental, it only leads to an incomplete conclusion, and hence, people must enrich themselves with knowledge of the situation and choose people from all sides of the population to get the whole or a wider picture.
Also read: Clevo NH70: The Future or The Past?
