A group’s baseline risk (or likelihood) of an event happening.
A short summary appearing at the top of a scientific paper that describes what the researchers did, what they found, and the implications of their results.
This part of a paper details how different researchers contributed to the study, and how the study was funded.
Information based on personal experiences or testimony, rather than scientific evidence, that might be presented as support for a particular claim
Specialized form of an observational study, in which researchers document a single situation—such as an individual patient’s unusual symptoms—in detail. Journalists should report on case studies with caution, since they can’t provide any broad conclusions.
Researchers publish commentaries to share their opinion on an issue or debate in their field. These are essentially scientific op-eds.
A range of values that likely includes the true value researchers are estimating in their study, determined by adding and subtracting the margin of error from the estimate. Large confidence intervals or overlapping intervals when comparing two groups are signs of less meaningful findings.
The tendency to search for, interpret, and process information in a way that supports our previously held beliefs.
Personal or financial ties that could potentially influence the results of a study or indicate a vested interest in supporting a certain scientific claim.
The group that does not receive the intervention or variable being tested in a study. In drug studies, the control group might receive a placebo, or inactive substance.
A relationship between variables. When two things increase or decrease together, they have a positive correlation. When one increases and the other decreases, they have a negative correlation. Correlation does not imply that one variable causes another.
The main point of contact for a paper. Often also the principal investigator (PI).
In this part of a paper, researchers summarize their findings and put them in context of their work and the field. This section often addresses any limitations, lingering questions, or potential directions for further research.
In this type of study, neither the participants nor the researchers know who is assigned to the experimental and control groups. Randomized controlled trials (RCT) are often double-blind.
A measure of the magnitude or strength of a finding, such as how different two experimental groups are. Small effect sizes indicate weak or potentially less meaningful findings. A finding can be statistically significant but have a small effect size
Information that can't be published until a certain date or time.
The group that receives the intervention or treatment being tested in a study.
The type of study in which researchers manipulate a variable, for example by introducing a new treatment or drug. These studies involve an experimental group, which receives the treatment, and a control group, which does not.
The element that differs between the experimental and control groups of a study. In a drug trial, the experimental variable would be the drug being tested.
A phenomenon that occurs when journalists, seeking to present all sides of an argument, give equal weight to both evidence-based claims and those without scientific backing.
Short for "Hypothesizing After the Results Are Known." This practice involves researchers retroactively generating a hypothesis to align with a statistically significant result that arises from running a large number of tests, or p-hacking.
A paper's section that includes the background infomation of a study such as the rationale, what previous research has found, and what questions the researchers wanted to answer.
Studies which follow participants over an extended period of time, often using a repeated measures design.
The degree of uncertainty around the researchers’ estimate, typically reported as a margin of error. For example, a poll might show one political candidate ahead of another 53 percent to 48 percent, ± 3 percent.
The average of a set of values. For example, if the number of students in each of 6 classrooms is 12, 18, 22, 24, 32, and 33, the mean is 23.5 (the total divided by 6).
The midpoint of a range of values. For example, if the number of students in each of 6 classrooms is 12, 18, 22, 24, 32, and 33, the median is 23 (half the classrooms have fewer than 23 students and half have more).
These papers pool data across multiple studies to bolster their tests with large sample sizes. Researchers conduct meta-anlyses to draw big-picture conclusions in a field.
The part of a scientific paper containing detailed, step-by-step descriptions of what the researchers actually did. Methods sections are written in such a way that a different team could do the same experiment in an attempt to replicate the results.
The practice of testing multiple combinations of variables with several statistical tests. Too many tests run without proper statsitical corrections can increase the chance of spurious findings, or those that are statistically significant just by chance alone.
In this study type, researchers gather data without manipulating a variable, such as in a review of public health data. These studies have no experimental or control groups but can be powerful for detecting patterns in large samples or measuring correlations between variables.
How likely an event is, divided by the likelihood that it won’t occur. Many scientific papers report their findings in this way.
Publicly accessible datasets that can be used and distributed freely.
Publications that provide free access to their articles.
The tools or tests that researchers use to assess relationships or effects in their studies. Researchers should specify which outcomes (also known as endpoints) they intend to measure before a study begins, and their published results should involve those same outcomes.
A number that’s extremely high or low, compared with the other values in a dataset. Researchers might exclude outliers from their analyses, potentially skewing their results.
Experts with relevant expertise who weren't directly involved in the research you're covering. These sources can offer important perspectives on a study's implications and weaknesses.
The questionable practice of running a large number of statistical tests on a dataset until a statistically significant result occurs. The more statistical tests researchers run, the greater their chance of finding a statistically significant result.
A measure of statistical significance, indicating the probability that the measured outcome would have occurred at random when there was actually no effect. In many fields, p <.05 indicates that a finding is statistically significant.
The process of subjecting a scientific paper to examination by a panel of experts in that field. The peer-review process is foundational to vetting research quality.
The relative difference between two percents. Percent change = (percentage point difference / starting value) x 100.
The difference between two percents when you subtract them. For example, if a flu outbreak infects 3 percent of the population one year and 5 percent the next, that’s a 2-point increase.
An inactive version of the treatment being tested in a study. This will sometimes be given to the control group.
An unpublished manuscript shared online by a researcher. Importantly, preprints haven't undergone peer review.
Often the senior scientist in charge of the lab and the team who ran the study. This person's name is typically listed last in a study's author list.
A belief or claim that's presented as scientific, without being supported by scientific evidence.
Communications staff at an institution who are responsible for liaising with media and providing information to both the press and the public. PIOs can help connect journalists with scientists who have relevant expertise for a story.
The phenomenon in science in which scientists or journals are more likely to publish studies with positive (statistically significant) findings over those with null (not statistically significant) or inconclusive findings.
A form of science which explores phenomena through non-numerical methods, such as focus groups, interviews, or detailed observation.
A form of science which investigates phenomena through numerical data and statistical analysis.
In this type of experimental study, experimental and control groups are randomly assigned to control for potential influencing factors. This study type is considered the gold standard in experimental design.
The list of other studies and publications cited in a study paper.
A comparison of the risk (or likelihood) of an event between two groups.
When scientists repeat a study (done by themselves or others) to see if they get the same results. If a finding holds across several studies, it's more likely to be true.
The practice of falsifying or fabricating results to increase the likelihood of publishing a finding or receiving research funding.
The paper section where researchers describe their findings in terms of the statistical tests they used and the statistical significance of those findings.
A publication providing a comprehensive overview of a field, often summarizing existing studies and identifying trends.
The number of people or other subjects (e.g., animals, cells, cities) being examined in a study.
Indicates the probability (denoted by a p-value) that a study's finding is a true effect, rather than having occurred just by chance alone. A p-value of < .05 (5 percent) is commonly considered to be statistically significant.
A separate, downloadable document containing additional information about a scientific paper's methods and results.
Outcome measures serving as sometimes-distant proxies for the actual outcomes being assessed. For example, a drug study might measure blood pressure as an indicator of heart disease risk.
A publication authored by a group of scientists or organization that takes a position on an issue, advocates for or against a policy, or offers some other sort of analysis. White papers aren't typically peer-reviewed.
Researchers conducting this kind of study (also known as a repeated-measures design) test the same participants multiple times and compare the results across time points. This design controls for individual differences, since each participant serves as their own control.
