Linear regression is a statistical procedure that allows the prediction of the values of a continuous dependent variable Y based on values of categorical or continuous independent variable X. In other words, this refers to how much of variance in Y can be best predicted by X. The relationship between X and Y is linear (which is usually first proposed theoretically). This linear relationship could be positive (i.e., an increase in X results in some increase in Y), or negative, which is vice versa or flat. When graphed, these predictions can help find the best fitted line—the line that touches as many points as possible—and hence the term, linear regression.