04 - Terminology and Bias
For a very decent list on what terminology means, visit the Australian Radiation Protection and Nuclear Safety Agency.
Some information about bias came from Phsychology Today. This will be Purple.
Some information about bias came from Cmotions. This will be Orange.
Some information about bias comes from Data36. This will be Green.
Some information about bias comes from SocialTalent. This will be Blue.
Bias - "Any systematic error in a study that that can lead to conclusions that are different from the truth."
Recall Bias - "Type of bias that occurs in epidemiological studies when self-reported, historical information is inaccurate. For example, in a case-control study cases may over-report past exposure, especially if it is widely known to be associated with the disease under study."
Selection Bias - "Type of bias that occurs from the manner in which subjects are selected for a study. When the sample of subjects is not representative of the target population the results of the study may not be valid. For example in a case-control study of smoking and lung cancer, the association between the two will tend to be weaker if the controls are selected from a hospital population (because smoking is associated with many other diseases resulting in hospitalisation) than if controls are selected from the general population"
Self-Serving Bias - "We attribute successes and positive outcomes to our doing, basking in our own glory when things go right; but, when we face failure and negative outcomes, we tend to attribute these events to other people or contextual factors outside ourselves."
The Curse of Knowledge - "Similar in ways to the Availability Heuristic (Tversky & Kahneman, 1974) and to some extent, The False Consensus Effect, once you (truly) understand a new piece of information, that piece of information is now available to you and often becomes seemingly obvious. It might be easy to forget that there was ever a time you didn’t know this information and so, you assume that others, like yourself, also know this information: the Curse of Knowledge. However, it is often an unfair assumption that others share the same knowledge."
Negativity Bias - "We like to win, but we hate to lose even more. So, when we make a decision, we generally think in terms of outcomes – either positive or negative. The bias comes into play when we irrationally weigh the potential for a negative outcome as more important than that of the positive outcome."
The Backfire Effect - "The Backfire Effect may work based on the same foundation as Declinism, in that we do not like change. It is also similar to Negativity Bias, in that we wish to avoid losing and other negative outcomes – in this case, one’s idea is being challenged or rejected (i.e. perceived as being made out to be ‘wrong’) and thus, they may hold on tighter to the idea than they had before. However, there are caveats to the Backfire Effect – for example, we also tend to abandon a belief if there's enough evidence against it with regard to specific facts."
The Fundamental Attribution Error - "The Fundamental Attribution Error is similar to the Self-Serving Bias, in that we look for contextual excuses for our failures, but generally blame other people or their characteristics for their failures"
In-Group Bias - "In-Group Bias refers to the unfair favouring of someone from one’s own group. You might think that you’re unbiased, impartial and fair, but we all succumb to this bias, having evolved to be this way. That is, from an evolutionary perspective, this bias can be considered an advantage – favouring and protecting those similar to you, particularly with respect to kinship and the promotion of one’s own line."
The Forer Effect - "Specifically, the Forer Effect refers to the tendency for people to accept vague and general personality descriptions as uniquely applicable to themselves without realizing that the same description could be applied to just about everyone else (Forer, 1949). For example, when people read their horoscope, even vague, general information can seem like it’s advising something relevant and specific to them."
Confirmation Bias - "Occurs when the person performing the data analysis wants to prove a predetermined assumption. They then keep looking in the data until this assumption can be proven. E.g. by intentionally excluding particular variables from the analysis. This often occurs when data analysts are briefed in advance to support a particular conclusion.
It is therefore advisable to not doggedly set out to prove a predefined conclusion, but rather to test presumed hypotheses in a targeted way."
Underfitting - "Underfitting means when a model gives an oversimplistic picture of reality. Overfitting is the opposite: i.e. when the model is overcomplicated. Overfitting risks causing a certain assumption to be treated as the truth whereas in practice it is actually not the case."
Confounding Variabelen - "Failing to allow for confounding variables can result in assuming there is a cause-effect relationship between two variables when there is in fact another variable behind the phenomenon. Bear in mind that a correlation is not the same thing as cause-effect."
Observer Bias - "Observer bias happens when the researcher subconsciously projects his/her expectations onto the research. It can come in many forms, such as (unintentionally) influencing participants (during interviews and surveys) or doing some serious cherry picking (focusing on the statistics that support our hypothesis rather than those that don’t.)"
Survivorship Bias - "Survivorship bias is a statistical bias type in which the researcher focuses only on that part of the data set that already went through some kind of pre-selection process – and missing those data-points, that fell off during this process (because they are not visible anymore)." E.G - A website's success stories won't include customers who aren't satisfied.
Omitted Variable Bias - "Omitted Variable Bias occurs when you are leaving out one or more important variables from your model."
Cause-Effect Bias - "Our brain is wired to see causation everywhere that correlation shows up.
Cause-effect bias is usually not mentioned as a classic statistical bias, but I wanted to include it on this list as many decision makers (business/marketing managers) are not aware of that. Even those who are aware of it (including me), have to remind themselves from time to time: correlation does not imply causation."
Conformity Bias - "If an individual feels the majority of the group are leaning towards/away from a certain candidate, they will tend to go along with the group thinks rather than voice their own opinions."
The Halo Effect - "Halo is when we see one great thing about a person and we let the halo glow of that significant thing affect our opinions of everything else about that person."
The Horns Effect - "The Horns effect is when we see one bad thing about a person and we let it cloud our opinions of their other attributes."
I'd like to add my own as I can't find a bias for this readily.
The alternative bias - Giving extra weight and belief in views that contradict the mainstream simply for it being different.
Control - "A sample or subject (animal/human) in which a parameter under investigation (cause or effect) is absent or is held constant, in order to provide a comparison. In an experimental study the experimental group is subjected to the factor under consideration, while the control group matches the experimental group in all aspects except that it is not subjected to the factor under investigation"
Causal - "Broadly, causality is the relationship between a cause and the consequential effect. In biology and medicine cause is the exposure and effect is the resulting disease or condition. It is important to realise that even if an association is found between an exposure and a disease the connection cannot automatically be interpreted as causal. An association can, in addition to being causal, be due to chance, bias or confounding."
Confounder - "An extraneous factor in a study which is related with both the probable cause and the outcome. A confounding factor may conceal an actual association or falsely demonstrate an apparent association between itself and the outcome where no real association between them exists. If confounding factors are not measured and considered, bias may result in the conclusion of the study."
Case Control Study - "A type of epidemiological study that compares subjects who have a disease or condition (the cases) with similar subjects who do not have the disease or condition (the controls). In a case-control study the medical and lifestyle histories of the subjects in each group are investigated to learn what factors may be associated with the disease or condition of interest."
Cohort Study - "A type of epidemiological study in which a particular outcome, such as a medical condition, is compared according to a putative factor (a factor suspected to influence the chances of acquiring the medical condition) in a group of individuals who are linked in some way (the cohort). In a prospective cohort study the group of individuals is followed over time in order to determine how the putative factor affects rates of the outcome of interest. In a retrospective cohort study, the data is collected from past records of the cohort."
Cross-Sectional Study - "A type of epidemiological study that aims to describe the relationship between a disease or outcome and other factors of interest as they exist in a subset of a population at a particular point in time. Since both the outcome and the factors are measured at the same point in time these studies are not strong at showing causal relationships."
Double-Blind Study - "An experimental procedure in which neither the subjects of the experiment nor the researchers know know the critical aspects of the experiment (e.g. who belongs to the control group or the experimental group). A double blind experiment is used to ensure impartiality, and avoid errors arising from bias."
Ecological Study - "A type of epidemiological study where data is analysed at the population or group level rather than the individual. They are inexpensive and easy to carry out, using routinely collected data, but they are prone to bias and confounding."
In Vitro Study - "Experimental study investigating biological properties on a subset of an organism’s constituent parts (e.g. organs, tissues, cells, biomolecules etc) and performed in a controlled environment such as in a test tube. Compared to in vivo studies in vitro studies are substantially faster, less expensive, and can be done with fewer ethical and safety concerns. However because the test conditions of an in vitro study may not correspond to the conditions inside of the organism, this may lead to results that do not represent the situation that arises in a living organism."
In Vivo Study - "Experimental study investigating biological properties on a whole living organism. In vivo studies are far more expensive, and often more difficult, than in vitro studies. However they are better suited for observing the overall effects of an experiment on a living organism."
Meta-Analysis - "A statistical analysis that combines the results of several studies that address a similar research topic. The advantage of a meta-analysis is that it combines all the research on the particular topic into one large study with many participants. The disadvantage is that the studies whose results are being combined may have different methodologies and experimental protocols (i.e. they are undertaken in a different way) and the results can become imprecise and difficult to interpret."
Pooled Analysis - "A statistical analysis similar to a meta-analysis where the results from a number of related studies (i.e. sharing a common protocol) are combined to provide an overall summary."
Population-Based Study - "A type of study that includes participants from the entire population of a defined area. For example, a study investigating mobile phone use and brain tumours amongst residents of Melbourne."
Provocation Study - "Experimental human study where subjects are exposed to either an agent that is claimed to provoke a response, or to a sham agent that should provoke no response."
Randomised Study - "An experimental procedure where the experimental units (e.g. the exposure under study) are randomly allocated across different groups. For example, if a provocation study is investigating the effects of radiofrequency (RF) radiation then the volunteers in the study are either exposed to RF or sham using randomization"
Retrospective Study - "A study that looks backward in time, usually using medical records (retrospective cohort study) and interviews with patients who already know to have a disease (case-control study). Retrospective studies have the benefit of being cheaper and less time consuming however sources of error due to confounding and bias are more common in retrospective studies compared to prospective studies"
Systematic Review - "A literature review that follows specific objectives, materials and methods, and conducted according to a specific and reproducible methodology (systematic way), to answer a specific research question."
Confidence Interval - "A range of values for a parameter of interest with a specified probability of including the true value of the parameter. Thus the confidence interval or CI is used to indicate the reliability of an estimate for the parameter of interest. The specified probability is called the confidence level, usually expressed as a percentage; thus one speaks of a "95% CI". Increasing the desired confidence level will widen the confidence interval."
Dose-Response - "The relationship between the amount of exposure (dose) to an agent and the resulting changes in the effect (response)."
Excess Relative Risk - "A statistical measure used in epidemiological studies to describe percentage change of an exposure causing a specific health outcome above the baseline risk.
For example, if a cohort study of radon exposure and lung cancer reported an excess relative risk of 0.16 per 100 Becquerel’s/m3 this would mean that there is an increased risk of 16% for the exposed population above the baseline risk."
Genotoxicity - "Harmful action on a cell's genetic material affecting its integrity. Genotoxic agents are known to be potentially carcinogenic (cancer causing), mutagenic (mutation causing) or teratogenic (birth defect-causing)."
Incidence - "Measure of the risk of developing a disease or condition during a specified period of time. Usually given as the incidence rate which is the number of new cases per population at risk during the specified time period. For example, over the period 2001-2005 the cancer incidence rate for children aged 0-14 years in Australia was 14 per 100,000"
Latency period - "The time elapsed between exposure to an agent and the clinical onset of a disease."
Nocebo - "An adverse health effect resulting from psychological factors due to a person's belief that something is harmful."
Placebo - "A beneficial health effect resulting from psychological factors due to a person's belief that something is beneficial."
Odds Ratio (OR) - "A statistical measure used in epidemiological studies to describe the strength of association between an exposure or factor and a particular disease or outcome. The odds ratio (OR) is a relative measure of risk, describing how likely someone who is exposed to the factor under study will develop the outcome as compared to someone who is not exposed. For example, a case-control study investigating smoking and lung cancer found an odds ratio of 17.4, i.e. the odds of smoking among the lung cancer patients was more than 17 times higher than the odds of smoking among the controls."
Risk Factor - "A risk factor is any attribute, characteristic or exposure associated with an increased risk of disease or adverse health outcome. A risk factor may not necessarily be causal."
Sham Exposure - "The environmental conditions of exposed samples/subjects, but in absence of the exposure. Usually associated with controls."
Specific Absorbsion Rate - "A measure of the rate at which electromagnetic energy is absorbed by the body when exposed to radiofrequency radiation. The specific absorption rate or SAR is defined as the power absorbed per unit mass of tissue and is measured in units of watts per kilogram (W/kg)."
Risk Ratio - "A statistical measure used in epidemiological studies to describe the risk of a disease or outcome relative to a particular exposure (Also called relative risk - Wikipedia). It is the ratio of the probability of an outcome occurring in a group exposed to a certain agent or factor versus a non-exposed group. For example, a cohort study investigating smoking and heart disease found a risk ratio of 5.2, i.e. smokers are five times more likely to develop heart disease compared to non-smokers."
That's it for now. Next wewill go over the basics of statistics to help you understand how to read scientific studies.
Next page: Science - Statistics