Predictor Variable

Basic Concepts and Terms

At the most fundamental level, predictor variables are variables that are linked with particular outcomes. As such, predictor variables are extensions of correlational statistics. Therefore, it is important to understand basic correlational concepts. As a function of the correlational relationship, when strength increases, slope, directional trends, and clustering effects become increasingly defined. Slope defines the degree of relationship between variables, with stronger relationships evidencing a steeper slope. The degree of slope varies depending on the relationship of the examined variables. Directional trends are either positive—as the x variable increases, the y variable increases—or negative—as either the x or the y variable increases, the other variable decreases. An example of a positive relationship is height and weight. As people grow, weight typically increases. The relationship is not necessarily a one-to-one ratio, and exceptions certainly exist; however, this relationship is frequently observed. The relationship between weight of clothing worn by people and temperature is an example of a negative relationship between variables. People are unlikely to wear fewer clothing items on days when the weather is below the freezing point, and they are just as unlikely to wear layers of clothing when the temperature nears 100°F. As mentioned in the previous example, variation will exist in the population—some will choose to wear a light jacket on warm days, whereas others may wear T-shirts and shorts. This variation is called clustering. Correlational clusters are the degree to which variables form near each other around an estimate line that provides a representation of a linear trend in the scatterplot. Researchers commonly refer to this linear estimate as the line of best fit. As the observed spread in the cluster decreases, the observed variables exhibit less variance from the line of best fit; this increases the precision of the linear estimate. In a perfect correlational relationship, no deviations from the line of best fit occur, although this is a rare phenomenon.

...