Biases in Research PDF

Biases in research Ivan Buljan Assistant professor  Content 0 1 Bias in research Experimenter bias Respondent bias  Expected outcomes 0...

Biases in research Ivan Buljan Assistant professor  Content 0 1 Bias in research Experimenter bias Respondent bias  Expected outcomes 0 2 Recognize common biases that can affect the design, data collection, and interpretation of research studies. Understand how biases can impact the validity and reliability of research findings. Learn strategies for preventing and mitigating biases in research Understand the ethical implications of bias in research  Systematic error/ bias 0 3 The results are systematically different than the real results in the population. This can happen in any research design and in any phase of research AND make the results not interpretable.  Systematic error/ bias 0 Researcher biases Respondents’ biases 4 Sampling bias Acquiescence bias Allocation bias Social desirability bias Non-response bias Habituation Information bias Sponsor (funder) bias Recording bias Confirmation bias Recall bias Culture bias Detection bias Question-order bias Drop out bias Leading question bias Ascertainment bias Halo effect Hawthorne effect Chronological bias  Most common types of bias 0 (researchers) 5 Selection bias Sampling bias (To determine the prevalence of a specific behavior, we invited participants via social media, and only those who were interested volunteered)  Selection bias 0 6 Allocation bias (In a study examining the effects of a new stress management program, the first 20 volunteers who signed up were assigned to the treatment group, while the next 20 were assigned to the control group. The allocation was not random.) Non-response bias (classical survey bias; in the survey research the response rate is usually 20%, and participants are more likely to be more educated, female, previously participated in survey research…)  Selection bias 0 7 Every sample which was not chosen at random can be a source of selection bias Here is the video about WEIRD research: https://www.youtube.com/watch?v=Ho6OlPrD7sA Randomization resolves problems related to selection bias in RCT. The error cannot be eliminated after the research is done and for that before researching one should: Define the inclusion and exclusion criteria Ensure a high response rate Know the natural course of the disease/phenomena  Biases 0 8 Information/measurement bias: it occurs if the instruments or the assessors are incorrect Recall bias: mothers of children with malformations will recall more details on risk factors compared to mothers of healthy children Schoth et al., 2020: A review that looked at how people with chronic pain remember pain-related information compared to neutral information, as well as how this affects their condition. 18 studies were found for this review. Some studies showed that people with chronic pain remembered painful words less when they weren't depressed, but this was not true for all pain-related words. There were no significant differences between groups in recalling words related to the sensation of pain, illness, or depression. People with chronic pain tended to remember sensory-pain words more than neutral or emotional pain words. They also remembered illness-related words more than depression-related words. In contrast, healthy individuals remembered neutral words better. However, the evidence for pain-related memory biases in chronic pain patients is unclear. More high-quality research is needed to understand this better.  Biases 0 9 Recording bias: In a study on social media usage and its impact on well-being, researchers collected data from online forums. They found that there was more information available on individuals experiencing social anxiety compared to those with strong social skills Detection bias: if groups were not diagnosed using the same procedure (e.g. depression diagnosed using different questionnaire)  Biases 1 0 Drop out and attrition bias: lost of participants through follow up; it can be source of bias, but it can also indicate which therapy is not suitable Ascertainment bias : refers to situations in which the way data is collected is more likely to include some members of a population than others. E.g. In a study on teenage smartphone addiction, using an online survey on a social media platform led to higher reported addiction rates among active users, indicating ascertainment bias and a potential overestimation of prevalence. This might not accurately represent the entire teenage population.  Biases 1 1 Hawthorne effect: if participants knew that they are being watched, that can affect their behaviour Chronological bias: changes in medical treatment and/or diagnosis of a disease over time. E.g. When examining the prevalence of ADHD in children across decades, shifts in diagnostic criteria and assessment tools may lead to higher reported rates in recent years compared to earlier decades. This doesn't necessarily represent a real increase in ADHD but underscores the impact of changes in diagnosis and assessment practices over time.  Biases 1 2 Lack of blinding: Anytime the respondents know they are being assessed, they will try to modify their behavior. Therefore blinding… BUT, not limited only to them… https://catalogofbias.org/biases/lack-of-blinding/  Respondents’ biases (Sarniak, 2015)1 3 Acquiescence bias: tendency of the participant to agree with moderator and be generally positive about anything moderator/research assistant says/does (“yea-saying”, friendliness bias) Social desirability bias: answering questions in a way that will lead to being accepted or liked  Respondents’ biases (Sarniak, 2015)1 4 Habituation: Respondents provide the same answers to questions that are formed in similar ways How does an individual's self-efficacy influence their motivation to pursue challenging goals? What role does self-efficacy play in shaping an individual's belief in their ability to overcome obstacles and setbacks? In what ways does self-efficacy impact a person's resilience when faced with adversity or failure? Sponsor (funder) bias: The feelings a person has towards sponsor may influence his/her answers.  Respondents’ biases (Sarniak, 2015)1 5 Confirmation bias: Researcher uses participants to confirm his/her hypothesis An individual who believes that a particular political party is corrupt might only follow news outlets and social media channels that highlight stories of corruption related to that party while ignoring stories that showcase their positive actions. This reinforces their existing belief and skews their perception of reality Culture bias: Assumptions we have on other cultures may bias our opinions. Also: Many standard IQ tests developed in Western countries include references and language that are familiar to Western participants but unfamiliar to those from non-Western cultures. This can result in lower scores for non- Western individuals not because they are less intelligent, but because they lack exposure to the cultural references used in the test items​  Biases beyond your control 1 6 “Hot stuff” bias- The current popularity of a topic can affect how much publicity is given to it. Publication bias- New and “attractive” significant results are more likely to get published  Bias: And another video  1 7 Publication bias: Publication bias occurs when some studies are published and others are not. Studies that remain unpublished tend to be systematically different from those that are published, meaning that conclusions drawn from the published literature will be systematically different from the truth (or biased). Positive trials and “exciting trials” are more likely to be published. https://www.ted.com/talks/ben_goldacre_battling_bad_science Psychology: https://www.psychologytoday.com/us/basics/replication-crisis  How to battle biases 1 8 Using Random Sampling: Ensures a more representative sample. Blinding: Keeps participants and experimenters unaware of the group assignments to prevent influence. Standardized Procedures: Use consistent methods across all participants. Peer Review and Replication: Helps identify and correct biases in research findings Preregistration of research: helps to remove biases before the study has even begun Thank you for your attention! Questions?