Association. A statistical relationship between two variables (groups or things). For example, people who golf have a lower risk of mortality. An important point is that an association is not a cause. An association only means two variables are related. The association may occur by chance or because there is an influence from these or other variables.

Baseline. The initial measurements of, or characteristics of, a group that will be followed during a study. Changes (or no changes) following the intervention or analysis will be compared to the data collected at baseline.

Blinded study. To avoid a conscious or unconscious influence on the study’s results, the participants are “blinded” so that they do not know who is receiving the treatment and who is receiving a placebo (a fake treatment). In a single-blind study, either the participant or the researcher does not know; in a double-blind experiment, both the participant and the researcher do not know who is receiving the treatment.

Causation. When one factor or action directly causes a result; a proven cause-and-effect. Causation is different from a correlation or an association.

Confounding. A confounder is a factor that confuses the association between two or more items and can lead to false conclusions. The confounder makes it difficult to discern if a result was because of the intervention or variables being studied. Confounding is more likely in observational studies that are not randomized.

Controlled trial. An experiment where people are assigned to a group to receive a treatment or intervention (for example, a vitamin, medicine or exercise program). The results are compared statistically with a similar group that did not receive any treatment or received a standard treatment (also called usual care) to evaluate the effect of the experimental intervention.

Correlation. A number that shows if there is a relationship between two quantifiable variables (for example, two lists of numbers) and how strong that relationship is. The correlation shows how close to a straight line the numbers fall. A strong correlation may or may not indicate that one variable causes the other.

Cross-sectional study. This study measures individuals, risk factors or another variable at one point in time. Because the data is obtained once, at one point in time, it represents a cross-section of the population being studied.

Disease burden. A statistical analysis that calculates the impact (the burden) of a health condition, injury or disease by creating an index of indicators, such as economic impact (e.g., days of lost work, medical costs), death, disability and loss of healthy life years. Risk factors may also be included in an estimate of disease burden.

Health-related quality of life (HRQoL). A measurement of how people view their physical and mental health, usually from answers to a questionnaire that are rated on a scale. The HRQoL measure is used in determining how a chronic or acute condition impacts an individual’s daily life.

Incidence. The number of new cases of a condition or disease at a point in time. Incidence is different than prevalence, which is the total number of people with a condition or characteristic at a given point in time.

Intervention. The treatment, program or action that is being tested in a study. An intervention can be as exercise program, behavioral strategy, vitamin regimen, education course or other type of treatment that is intended to “intervene” or change the course of an outcome.

Longitudinal study. The people or items are followed over time, usually with one or more periods when the variables are measured. Longitudinal studies take place over years, and are effective in showing changes related to the aging process. A longitudinal study differs from a cross-sectional study, which collects data solely at a single point in time.

Margin of error. When a study or poll samples a portion of a population (since it is impossible to study, for example, the entire population of New York City), the margin of error is a statistical estimate of how well the results reflect the attitudes or beliefs of the entire population.

Meta-analysis. Statistical method to combine the results of multiple studies to create a large sample size and a new summary of results. Studies in a meta-analysis must contain data that can be compared.

Observational study. A group of people is followed, or observed, over time and the outcomes that are being measured are recorded. There is no intervention, treatment or other attempt to affect the outcomes. Some observational studies continue over decades.

Omnibus survey. The organization fielding the survey provides a group of respondents with the desired demographic characteristics, and sends a survey on multiple topics with questions supplied by organizations that pay to have their questions included. The paying organizations “ride on the omnibus.” Companies share respondent demographics, but receive only the responses to their own questions.

Opinion poll. Opinion research collects the viewpoints of a group of people who answer a set of questions. Polls can be scientific, asking a sample of the population selected by the researcher so that results represent the larger population, or nonscientific, asking a group of people who may be volunteers or biased for or against the survey questions. The wording of poll questions also can bias responses.

Pilot study. A small study to explore the feasibility of conducting a larger study. Pilot studies do not answer the research question, but they do provide details on how a full-scale study could be conducted and analyzed, costs and practice in using a methodology.

Prevalence. The number of people with a condition or disease at a give point in time. [Prevalence is different than incidence, which is the number of new cases of a condition or disease at a point in time.]

Projections. Projections are estimates, or “what if” scenarios, of changes that could take place in the future based on current information and past trends. Projections are based on a set of assumptions and often propose at least three scenarios: low, midrange and high estimates. When the assumptions or the current information changes, the projection will be affected.

Qualitative research. A subjective method that identifies and analyzes people’s beliefs, attitudes and opinions to discover the “quality” of a topic. Often using structured questionnaires, answers are organized and analyzed for themes, trends and specific comments to gain insight into a topic.

Quality Adjusted Life Year (QALY). The QALY calculation combines the quantity (lifespan) and quality of life, with 1.0 being one year of perfect health and death being 0.0. A year of life with a disability might be 0.6. QALY is used to estimate disease burden and the effectiveness of public health interventions and health care.

Quantitative research. A research method that uses objective measurements and numbers to examine a research question. Quantitative research is objective and enables statistical analysis.

Random sample. People (the “sample”) are chosen by chance. The people in the sample may answer survey questions, or be assigned to a group that will test different interventions in a study. Randomness insures that the makeup of the group or the outcomes of an intervention cannot be predicted. The purpose of a random sample is to be able to generalize the results to a larger group of people.

Randomized controlled trial. Participants are randomly assigned (by chance) to one or more experimental intervention/treatment groups or a control group. The results of the experimental treatments are compared to the results of the control group, which receives a placebo, an accepted standard treatment or no treatment at all.

Relative risk. A comparison between two things, with the differences stated relative to one another. For example, people who have a risk factor compared to people who do not have it. Relative risk is usually expressed with phrases such as “two times greater risk,” “decreased by 11%” or “three-fold greater risk.”

Representative sample. The sample (group that is being studied) is as close as possible to the characteristics of the larger group it is pulled from. In a study of people, the sample may be selected to represent characteristics such as health behaviors, habits or beliefs, and the demographics may be selected to be similar to the proportion of men, women, ages and ethnicities found in the larger population.

Retrospective study. Existing data from past studies or medical records is re-examined. Retrospective means looking back in time. The benefit is that the data is already available; the caution is that the quality of the data relies on the accuracy of the collection in a past time, which the researcher cannot control.

Risk factor. A trait, genetic characteristic or lifestyle choice that increases the probability of an injury or a disease, identified by a statistical association between the risk factor and people who experience the injury or disease. A risk factor does not mean a person will get the disease, but it is more likely.

Significance. Statistical significance means it is not probable that the results are due to chance alone. A higher level of significance increases the confidence that the results are “true.” Significance is a function of the number of participants and what is being measured.