selection
Research is an activity that needs to be shared as a whole picture to the world. If some parts are incomplete or misleading, it can lead to wrong conclusions. When errors like selection bias influence a reseracher, it can lead to people understanding the wrong thing. But what is it? And why is it a big deal? This article will explain with examples and other details.
What is Selection Bias?
Selection bias is an error that occurs in research when the respondents in the experiment or survey do not represent the entirety of the population. It’s like trying to show the entire picture by using one piece of the puzzle, which can be misleading, as it only shows a portion of the whole image, and some groups may be underrepresented.
How does it happen?
Selection bias may be an unintentional oversight or intentional due to personal bias or other reasons. Here are some explanations as to why it could occur:
- Selection process: During the selection of the sample population, researchers may select people from a single public space, which can lead to gathering people who belong to various parts of the population, but not all of them. Researchers may choose the nearest people for the sake of convenience and finishing up any field work faster (which is also known as convenience sampling)
- Recruitment: If the participants are selected through volunteering, then only interested and motivated people are more likely to be part of the activity. This leaves out people who do not wish to participate but want to tip the scales with their opinions. They may be more responsive in an online survey.
- Follow-up: Many people may be eager to participate in a survey, but not all would be willing to stay if it is long-term. They may drop out for many reasons (like bad experiences or suddenly being occupied with something else), and this creates a difference in the study. The research, which was conducted with a certain number of people, won’t end with the same number. And the latter part of the research may include important aspects about the intended objective, and hence, it won’t be balanced.
- Data-access issues: The researcher may use pre-existing models that consist of forms of selection bias. Like using hospital surveys, where only those who attend the hospital are involved.
- Geographical and temporal restrictions: The researcher may choose the wrong place and time for fieldwork. Like times where most people are exhausted and may be interested but won’t be able to say it.
Types of Selection Bias:
Here are the types of selection bias that could be observed in research:
- Sampling bias: It occurs when the people chosen for the survey or experiment do not represent the entire population. For example, choosing only tea and coffee lovers to represent the entire population who enjoys ‘beverages’, as many drinks also fall into that category.
- Non-response bias: It occurs when the sample population is not active in the survey. It could be for many reasons, and a few of them could be bad experiences in the past or other personal reasons.
- Self-selection bias: This error occurs when people volunteer for the survey. Though not a thing to be discouraged, the people who join the experiment due to their interest may have opinions and experiences that vary significantly from those who do not.
- Exclusion bias: This mistake occurs when researchers exclude certain subgroups that they think do not meet their criteria. Like trying to conduct research on people in Spain about their experiences, but excluding the foreigners who had settled there for years.
- Attrition bias: This happens when people do not wish to continue participating in the experiment until the end, and is also known as the Loss to Follow-up bias.
- Healthy user bias: This occurs when only the healthy individuals are considered for a survey. Or when only the positive and successful individuals are selected, but not the ones who dropped out in between for personal reasons.
- Berkson’s bias: This happens when the sample population is hospitalised individuals. This can create a situation where a symptom suffered because of another condition can be mistaken as a symptom of another disease, which can lead to wrong conclusions.
- Surveillance bias: Also known as Detection bias, it occurs when researchers only observe a group of the population closely as to increase the chances of detecting their objective, not considering variables. It also concludes that certain conditions are more frequent in that group, which may not be true.
Case studies:
Here are some case studies where selection bias was prevalent:
- Hormone Replacement Therapy and Heart Disease (1990s):
In this case, studies showed that females who took Hormone Replacement Therapy (HRT) suffered from less chances of suffering from any heart-related ailments. But it was later shown that the sample population belonged to circles who had proper access to healthcare, were health-conscious, and the wealthy. Those conditions automatically allow them to have good health, and hence, that was a major factor for the low chances of heart diseases, not the HRT alone. This form is Self-selection bias or Healthy user bias. Later, it was deeply studied by a randomized controlled trial (RCT) by the Women’s Health Initiative, which found that HRT not only had no particular influence on cardiovascular health, but also had potential risks.
- COVID-19 Detection through App (2020):
Some countries, like the UK (NHS COVID-19 app), made an application that aimed to detect infection rates of the virus during the Pandemic. But its outreach was confined to only the ones who had a smartphone, among other factors. Hence, the ones who didn’t have smart gadgets, like the homeless and rural population, could not be included, and the results were inaccurate. This is counted under self-selection bias and Digital access bias.
- Loan Approval:
Banks built models that studied only the ones who were selected and approved for loans, but not the ones who were rejected. This can lead to wrong conclusions, and the predictions could not be that accurate.
Impact of Selection Bias:
Selection bias can be a blind spot for many researchers and can lead to wrong results or misinformation. It can impact the research and opinions that are inspired by it, and thus help the dissemination of the bias.
- Impact in Psychology: If the sample population consists only of college students, assessing the objective in adult and older generations can be hard. Hence, choosing a single age group for research that represents all ages isn’t right, especially when several factors influence a person as they age. Conditions for one age group cannot be used to generalise for the entirety. Theories experimented on using a specific population are incomplete, as they may also affect the rest of the population groups.
- Impact on Business: If the company solely relies on customer surveys that were answered by the eager and active, they may not know the complete situation around their products or services. Some may have complaints, but for one reason or another, they may not answer or access the survey, and the company might not be able to find areas to improve their product. And if they rely on people who have been using their products for a long time, they might undermine the new customers.
- Impact on Education: If educational institutions used the scores of well-performing students, it may not represent aspects like teacher methodologies, as the students might have gotten impressive scores without the latter due to reasons like hard work and a good foundation in academics. Hence, choosing the best of the best to represent the goodness of the school may not be a good idea, and it is quite common in many countries. Students may assume that marks are everything in life and may not focus on social and interpersonal skills.
- Economic policies: The influence of selection bias in framing economic policies can have everlasting consequences. Unless it is a specific objective, having a sample population that doesn’t represent the entirety and using it to frame crucial policies for the whole country simply cannot meet all needs. While doing a public survey, people from all parts of the country need to be considered, and such decisions cannot be centralised, as every region may have a different requirement or aspect to develop.
- Healthcare: If large-scale clinical trials exclude subgroups like pregnant women and elderly people for the sake of convenience, it cannot paint the whole picture. And if the hospital focuses only on its patients for observation, it might undermine the severity of the situation. This specific exclusivity may result in ineffective medications. If a medicine is marketed for everyone while not being tested on everyone, that breeds more problems later on and may even end in drastic results.
How to detect it?
Selection bias is more complex to alter, especially after the research is complete. Hence, it justifies the quote “prevention is better than cure”. Here are methods to detect it:
- Examination: Check if the sample population consists of people who represent every corner of the population.
- Reflection: Analyze the methods and criteria you used to select the sample population. And question why you excluded certain subgroups.
- Compare groups: Check the characteristics of everyone in the group.
How to avoid it?
- Use Random Sampling: Choose people randomly, without any criteria in mind. Use methods like systematic and strategic sampling, where the subgroups are created neatly and can include all people that represent the population as a whole.
- Track participation: Track the progress and also include the ones who did not participate actively or dropped out in between.
- Tests: Use your findings and try to observe them in a different group to see if the results are recurring.
- Self-selection: Not only choose people who volunteer, but they also provide incentives and different forms of surveys (like online forms) to gather their opinions too.
To conclude, Selection bias must be avoided to get accurate results. Inclusivity is very significant in making theories and decisions that are intended to apply to the whole population. Selection bias ignores and underepresents certain subgroups, whose opinions can add weight to research. Hence, by actively trying to avoid it and gathering opinions of all representatives of the population to craft into their findings, researchers can find accuracy and fewer complications.
Write and Win: Participate in Creative writing Contest & International Essay Contest and win fabulous prizes.
About the Author
