Selection bias refers to samples taken for research which do not reflect the rest of the population accurately, such as performing a study of obesity among children but only using middle class children from one area to represent all of the United States. Selection bias also occurs when two groups with significant differences are used in the same study; if a researcher performs an experiment into the effects of stress at the work place in comparison to stress at home, two similar groups should be used for each situation to avoid creating two significantly different results because of group differences. This may also refer to studies or materials which are included in the research, such as only citing previous studies which support the research outcome.
Measurement bias includes using instruments and research methods which will effect the overall results or are not ideal for the type of research, research tools that are not sensitive enough to detect small variables which could impact results (insensitive measure bias) or using instruments that may influence results towards the researcher's outcome (expectation bias). Researchers who are measurement-biased may use methods that will influence their outcome or support their initial predictions, such as creating questionnaires where the questions may influence the response. In some cases, measurement bias can be entirely accidental, such as using unbalanced weighing scales in experiments.
Intervention bias refers to how much the researcher, or other factors, intervene with the test subject(s). For example, one group in a clinical research trial may be told that they will be receiving a placebo drug and may act differently due to this knowledge (co-intervention bias). Intervention bias may also refer to the timing of research. For example, would results have varied if the performed research took place over three weeks instead of three days? Or it may occur when intervention or treatments are not applied equally to all groups.
Bias can also occur in organizations who have issued a research report or organizations funding the research. If the researcher produces a report that is generally supportive of their sponsoring organization, or allows the design and methodology of her research to be influenced by the organization that has provided funds for the research, then the research is considered funding-biased. Similarly, research reports designed to be accepted by certain publishers or if the publisher has a preference for the report solely on the basis of its findings may be considered publication-biased.