15.13
Correlations
Correlation means that there is a relationship between two or more variables (such as ice cream consumption and crime), but this relationship does not…
Sometimes, researchers may choose to more passively interact with a phenomenon, rather than to intervene and manipulate the behaviors of interest.
For instance, perhaps a scientist wants to know if there’s an association between eating a plant-based diet and sleep. In this case, she quantitatively measures two variables by asking participants to report how many vegetables they consumed that day, and later, how many hours they slept.
This design is known as correlational research—examining whether relationships exist between two variables.
After collecting both series of measurements, the researcher can visualize the data for each unit, in this case each person, on a graph—a scatterplot—with the variables—daily vegetable consumption and amount of sleep—placed on either axis.
Statistically, correlations are determined by calculating the correlation coefficient, commonly denoted as r—a number between -1 and 1 that indicates the direction, the sign, of the association and its overall strength, how tight the points align.
Here, the correlation could be positive, which means that the two variables move in the same direction. That is, people who consumed very few veggies slept less, whereas others who ate more, slept more.
In addition, the correlation could be strong, with a value close to 1, which indicates that the data points cluster linearly, with very few exceptions across individuals.
When the association has more exceptions, the linear pattern can disappear as the data points are spread out: the absolute value becomes closer to zero, and the correlation is considered weak to non-existent.
Now, if the two variables move in the opposite direction of each other, the correlation would be considered negative. That is, the more vegetables people consumed, the less they slept and vice versa. The strength is relatively strong, given the scatter is rather linear.
Importantly, correlation does not mean causation! Researchers would need to follow-up on observations to determine just how changes in one variable cause changes in another.
View the full transcript and gain access to JoVE Core videos
Q1: What is a correlation coefficient and what does the r value tell us?
A correlation coefficient, denoted as r, is a number between -1 and 1 that indicates both the direction and strength of the relationship between two variables. The sign shows direction: positive means variables move together, negative means they move oppositely. The absolute value's closeness to 1 indicates strength; values near 1 show tight linear relationships, while values near zero indicate weak or nonexistent associations.
Q2: How do positive and negative correlations differ in practice?
In a positive correlation, variables move in the same direction—as one increases, the other increases too. For example, vegetable consumption and sleep hours may correlate positively. In a negative correlation, variables move oppositely; as one increases, the other decreases. The strength of either type is determined by how closely data points cluster linearly around a trend line.
Q3: Why is correlation not the same as causation?
Correlation describes a relationship between variables but does not prove that one causes the other. A third variable may actually drive both observed changes. For instance, wealth correlates with intelligence, but education—a third variable—may be the true cause. Researchers must conduct follow-up studies to establish cause and effect relationships between variables.
Q4: What does it mean when a correlation coefficient is close to zero?
When a correlation coefficient approaches zero, the relationship between two variables is weak or nonexistent. Data points scatter widely rather than clustering along a linear pattern, making it difficult to predict changes in one variable based on the other. For example, hours of sleep and shoe size would likely show a correlation near zero.
Q5: How can researchers use correlations to make predictions?
Correlations have predictive value when they are strong. For example, university admissions committees can correlate current students' college GPA with standardized test scores to predict success of applicants. The stronger the correlation observed in existing data, the more reliably researchers can forecast outcomes for new cases, though prediction accuracy decreases as correlation strength weakens.
Q6: What is a scatterplot and how does it help visualize correlations?
A scatterplot is a graph that displays data points for each unit studied, with two variables placed on separate axes. It allows researchers to visually inspect the relationship between variables. When points cluster tightly along a line, correlation is strong; when scattered widely, correlation is weak. The scatterplot's pattern reveals both direction and strength at a glance.
Q7: How does correlational research differ from experimental research methods?
Correlational research passively measures existing relationships between variables without manipulation, such as surveying vegetable consumption and sleep hours. Experimental research actively intervenes by manipulating variables to test cause and effect. Correlational designs are less invasive and useful for discovering associations, but cannot establish causation like experimental designs can.