Bias + Error

Bias: Deviation of results/inferences from the truth, or processes that lead to systematic error. This can be a result of prejudice for or against something. Since bias is systematic, large sampling will not eliminate the deviation. Attrition Bias: Type of selection bias due to systematic differences between study groups in the number and the way participants are lost form a study. Base Rate Fallacy: People tend to ignore the base rate (general prevalence) in favor of individuating info (info for only one case). Berkson‘s Bias: Selection bias that causes hospital cases & controls in a case control study to be systematically different from one another because the combination of exposure to risk & occurrence of disease increases the likelihood of being admitted to the hospital. Confounding: An extraneous factor (confounder) is related to both exposure & outcome, causing a distortion in the association. To reduce, restrict study group by maintaining the same factors in study, match comparison groups with a single confounding factor, or randomization. (I.e., Rates of eating ice cream & drowning both go up in the summer). Ecological Fallacy: Logical error that occurs when the characteristics of a group are attributed to an individual. In other words, ecological fallacies assume what is true for a population is true for the individual members of that population. Hawthorne Effect: When the individuals of a study will modify their behavior due to their awareness that they are being studied. Healthy Worker Effect: Those who are not healthy are less likely to be employed, employees who are, tend to be less sick. As a result, comparisons of mortality rates between an employed group & the general population will be biased. Information Bias: Distortion in the measure of association caused by a lack of accurate measurements of key study variables. Results from systematic differences in the way data on exposure/outcome are obtained from various study groups. Interviewer/Recorder Bias: Systematic difference in soliciting, recording, or interpreting information on exposure/outcome or when reviewer interprets or records information differently for one group or if the reviewer searches more diligently for information in one group. Measurement Bias/Systematic Error: Errors in measuring data on exposure/outcome results in deviation from truth. (Ex:A scale always adds 2 lbs to actual weight). Lead Time Bias: When patients that are diagnosed earlier in a study seem to live longer as they have had symptoms for less time before diagnosis compared to others. Length Time Bias: Overestimation of survival duration due to the relative excess of cases detected that are asymptomatically slowly progressing, while fast progressing cases are detected after giving symptoms. Loss to Follow Up Bias: Subjects are lost over time in an uneven amount between the different groups (i.e. 40% of cases are lost but only 10% of controls are lost). Panel Effect/Error: When subjects are put through many interviews, making them tired. Publicity Bias: Bias of journalists and news producers within the mass media in the selection of many events and stories that are reported and how they are covered. Recall Bias: Systematic differences in the way subjects remember or report exposures/outcomes. Selection Bias: A bias caused when subjects are selected based on a third variable, meaning that the sample group is not an accurate representation of the target population. Simpson’s Paradox: Statistical paradox where it is possible to draw two opposite conclusions from the same data depending on how you divide groups (combine them, split them up, etc.). Must look at other causations (i.e. unequal distributions of different groups). Social-Desirability Bias: A type of response bias that occurs when survey respondents provide answers according to society’s expectations, rather than their own beliefs or experiences. Volunteer Bias: Those who volunteer to participate in studies are different than those who don’t. Non-differential Misclassification: When case definition isn’t well defined, so there’s mixing between controls and cases. Error: Difference between the true value and the measured value. Random Error: Divergence from the truth due to random chance. No preferred direction, so averaging over a large number of observations will yield a zero net effect. Statistical Errors: Type I Error: Rejection of a true null hypothesis/acceptance of a false alternative hypothesis (false positive). Type II Error: Failure to reject a false null hypothesis/rejection of a true alternative hypothesis (false negative). Type III Error: Correctly reject the null hypothesis for the wrong reason. Avoided by doing a two tailed test. Type IV Error: Correctly reject the null hypothesis but misinterpret results. Selection Bias happens during participant recruitment or selection, leading to a non-representative sample of the target population, which distorts the external validity (generalizability) of the results. Ascertainment Bias occurs during data collection, where there are systematic differences in how data is obtained or outcomes are assessed across different study groups, affecting the internal validity of the study (the accuracy of the conclusions based on the data). Information Bias is about errors in the data itself, whether due to inaccurate measurement, misreporting, or other inaccuracies in how data is collected or recorded.