The following regression equation explains the relationship between the dependent variable and independent variable:

Y_i = b₀ + b₁X_i + e_i, i = 1, 2, ..., n

There are two regression coefficients in this equation:

• The Y-intercept is given by b₀; this is the value of Y when X = 0. It is the point at which the line cuts through the Y-axis.
• The slope coefficient, b₁, measures the amount of change in Y for every one unit increase in X.

Also note the error term, denoted by e_i. Linear regression is a technique that finds the best straight-line fit to a set of data. In general, the regression line does not touch all the scatter points, or even many; the error terms represent the differences between the actual value of Y and the regression estimate.

Suppose a regression formula is, y_i = -23846 + 0.6942 x x_i, where the large constant and the x terms are on a date scale for a spreadsheet program (e.g., Sept 1, 1997 = 35674).

In the regression equation above, the b₁ term, 0.6942, represents the slope of the regression line. The slope, of course, represents the ratio of the vertical rise in the line relative to the horizontal "run." The slope and the intercept (b₀, here -23846) represent parameters calculated by the regression process. The process of linear regression calculates the parameters as those that minimize the squared deviations of the actual data values from the estimated values obtained using the regression equation. The process of determining these parameters involves calculus.

Linear regression involves finding a straight line that fits the scatter plot best. To determine the slope and intercept of the regression line, one must either use a statistical calculator, a software program, or calculate it by hand. Typically, an analyst will not have to calculate regression coefficients by hand.

• For regression equations with only one independent variable, the estimated slope, b₁, will equal the covariance of X and Y divided by the variance of X.
  b₁ = Cov(X,Y) / (σ_x)² = Cov(X,Y)/Var(X)

• The intercept parameter, b₀, can then be determined by using the average values of X and Y in the regression equation and solving for b₀.
  

The "hat" above the b terms means that the value is an estimate (in this case, from the regression).

The equation that results from the linear regression is called linear equation, or the regression line.

• Regression line: Y-bar_i = b₀(hat) + b₁(hat) X_i
• Regression equation: Y_i = b₀ + b₁X_i + e_i, i = 1, 2, ..., n

The linear equation is derived by applying a mathematical method called the least squares regression, also known as the sum of the squared errors: the sum of the squared distances of all the points in the observation away from the mean.

Note that analysts never observe the actual parameter values b₀ and b₁ in a regression model. Instead they observe only the estimated values b₀ and b₁. All of their prediction and testing must be based on the estimated values of the parameters rather than their actual values.

After you calculate the b₀ and b₁, test their significance. Also note that the regression analysis does not prove anything; there could be other unknown variables that affect both X and Y.