Explanatory versus Response Variables
Explanatory versus Response Variables
In a statistical study that uses one quantitative variable to predict or explain another, the predictor is the explanatory variable and the predicted quantity is the response variable.
The explanatory variable is also called the predictor or independent variable, traditionally denoted x; the response variable is also called the dependent variable, denoted y.
Not every study of two quantitative variables assigns these roles: one that only asks how strongly two quantities move together, without predicting or explaining either from the other, has no explanatory and no response variable.
Where the roles do apply, which variable plays which is decided by the purpose of the study; it is not a fixed property of the data themselves.
In the simple linear model
the symbol x is always the explanatory variable and y is always the response variable. By convention the explanatory variable is plotted on the horizontal axis and the response variable on the vertical axis, so that the picture matches the equation: read across to a value of x, then up to the corresponding y.
Weekly advertising spend and sales revenue for a retail store
Worked example
A physiotherapist records, for each patient in a clinic, the number of minutes of prescribed exercise completed in a week and that patient's reported pain score at the end of the week.
The clinical question is whether exercise time can be used to anticipate pain scores. Because the quantity to be predicted is the pain score, exercise minutes is the explanatory variable and pain score is the response variable.
On a scatterplot, exercise minutes goes on the horizontal axis and pain score on the vertical axis.
Left: the scatterplot convention, with the explanatory variable on the horizontal axis and the response variable on the vertical axis. Right: the explanatory variable is the input used to predict the response variable.
Notice what fixed the roles: not which variable was easier to measure, not which was recorded first, and not any claim about cause. The deciding question was simply which quantity the study wants to predict.
Counting explanatory variables
A study may use several predictors at once. If a model predicts one outcome from three measured characteristics, there are three explanatory variables and one response variable. The outcome being predicted is never counted among the predictors, even though it was measured in the same data collection.
A nutrition study records, for each of 50 participants: daily fibre intake in grams, daily sugar intake in grams, age in years, and cholesterol level in milligrams per decilitre. The researchers want to build a model that predicts cholesterol level from the other three measurements. How many explanatory variables does that model use?
Give your answer in variables.
The chart shows weekly advertising spend and weekly sales revenue for a retail store, collected so that the manager can forecast sales from a planned advertising budget. Based on the chart and the stated purpose, which variable is the response variable?
Left: the scatterplot convention, with the explanatory variable on the horizontal axis and the response variable on the vertical axis. Right: the explanatory variable is the input used to predict the response variable.
Weekly advertising spend and sales revenue for a retail store