Bias > Publication Bias
What is Publication Bias?
Publication bias is when studies with positive findings are more likely to be published — and they tend to be published faster — than studies with negative findings. This means that any meta analysis or literature reviews based only on published data will be biased, so researchers should make sure to include unpublished reports in their data as well.
Published vs. Unpublished Studies
“Published” means that the study has been published in a peer-reviewed journal. Studies are more likely to be published if they have positive findings, build on previously accepted hypotheses, and can potentially garner citations for the journal (e.g. if they have sensational findings). Studies are much less likely to be published if they don’t build on previously published data or if they refute a previously published hypothesis.
Around 50% of studies are estimated to be unpublished. In general, those studies are more likely to have less significant or negative results; that doesn’t mean the results aren’t valid — just that journals are less likely to publish an article or delay publication if a treatment is shown to have no effect. For example, a major study which showed a deworming program in India was not effective for reducing mortality or improving weight gain was delayed from publication for 8 years (Hawkes).
This “swept under the rug” phenomeneon happens as a result of withholding negative results from publication. This may be intentional or unintentional. As well as fraud, study sponsors may provide incentives in a deliberate attempt to skew findings. Journal editors might be more inclined to publish studies that will sell copy or reap other rewards.
Publication bias refers to an entire study being excluded. Similar biases include:
- Citation bias: finding literature sources by scanning reference lists from published articles. Less references sources are therefore more likely to be excluded from a meta analysis.
- Dissemination bias: when the nature of a study’s direction or the study’s results are unevenly reported.
- Gray-literature bias: ignoring literature that’s harder to find, like government reports or unpublished clinical studies.
- Language bias: the exclusion of foreign language studies from your analysis.
- Media attention bias: studies that show up in the news are more likely to be included in analyses than those that do not.
- Outcome-reporting bias: when positive outcomes are more likely to be included in a meta analysis than negative outcomes. Negative outcomes can also be misrepresented as positive ones.
- Time-lag bias: studies with significant results have a shorter median time to publication (4.7 years) while those with non-significant results have a median time of 8.0. years (Cochrane.org).
Hawkes N. Deworming debunked. BMJ. 2013;346:e8558.
Song F, Parekh S, Hooper L, et al. Dissemination and publication of research findings: an updated review of related biases. Health Technol Assess. 2010;14(8):iii, ix–xi, 1–193.
If you prefer an online interactive environment to learn R and statistics, this free R Tutorial by Datacamp is a great way to get started. If you're are somewhat comfortable with R and are interested in going deeper into Statistics, try this Statistics with R track.Comments? Need to post a correction? Please post on our Facebook page.