The least-squares straight-line slope is the coefficient that minimizes the sum of squared vertical residuals between observed data points and a fitted linear model. It measures the estimated change in the dependent variable for each unit change in the independent variable.
Least-Squares Straight-Line Slope
The least-squares straight-line slope is the coefficient that minimizes the sum of squared vertical residuals between observed data points and a fitted linear model. It measures the estimated change in the dependent variable for each unit change in the independent variable.
\(b_{1} = \sum_{i=1}^{n}(x_{i} - \bar{x})(y_{i} - ȳ) / \sum_{i=1}^{n}(x_{i} - \bar{x})^{2}\)
What this formula means
Scientific principle
Principle
Least-squares linear regression fits a straight line by minimizing the sum of squared differences between observed dependent-variable values and the values predicted by the line. The slope is determined by the covariance between x and y divided by the variance of x.
Derivation
1. Assume a linear model: yᵢ = b₀ + b₁xᵢ + εᵢ.
2. Define the residual for observation i as rᵢ = yᵢ − (b₀ + b₁xᵢ).
3. Minimize the residual sum of squares: SSE = Σrᵢ².
4. Differentiate SSE with respect to b₀ and b₁ and set both derivatives equal to zero.
5. Solving the resulting normal equations gives b₀ = ȳ − b₁x̄.
6. Substituting the intercept relation into the slope equation gives b₁ = Σ(xᵢ − x̄)(yᵢ − ȳ)/Σ(xᵢ − x̄)².
Final Result
b₁ = S_xy/S_xx
Slope Definition
S_xy = Σ(xᵢ − x̄)(yᵢ − ȳ)
Independent Variable Variation
S_xx = Σ(xᵢ − x̄)²
Intercept Relation
b₀ = ȳ − b₁x̄
Fitted Line
ŷ = b₀ + b₁x
Variables and their meanings
| Symbol | Meaning | Unit |
|---|---|---|
b₁ | Least-squares slope | Unit of y divided by unit of x |
b₀ | Least-squares intercept | Same unit as y |
xᵢ | ith observed independent-variable value | Unit of x |
yᵢ | ith observed dependent-variable value | Unit of y |
x̄ | Arithmetic mean of the independent-variable observations | Same unit as x |
ȳ | Arithmetic mean of the dependent-variable observations | Same unit as y |
S_xy | Sum of cross-deviations between x and y | Product of the units of x and y |
S_xx | Sum of squared deviations in x | Square of the unit of x |
n | Number of paired observations | Observations |
SI units and conversions
Slope
Dependent-variable unit per independent-variable unit
Intercept
Same unit as dependent variable
Residual
Same unit as dependent variable
Sum Of Squared Residuals
Square of dependent-variable unit
Conditions of use
There must be at least two paired observations.
The independent-variable observations must not all be identical, so S_xx must be nonzero.
The model assumes a straight-line relationship is suitable for the data.
Ordinary least squares minimizes squared vertical residuals in the dependent variable.
For conventional statistical inference, residual assumptions such as independence and appropriate variance structure must also be considered.
Limitations and assumptions
Ordinary least squares treats the x-values as fixed or measured with negligible error.
Substantial uncertainty in x requires an errors-in-variables or other suitable regression model.
Outliers can strongly influence the fitted slope.
A high-quality linear fit does not establish causation.
The fitted slope can be misleading if the relationship is nonlinear or if important variables are omitted.
Step-by-step worked examples
| Problem | Solution | Answer |
|---|---|---|
Fit the least-squares straight-line slope for the data points (1, 2), (2, 4), and (3, 5). |
| The least-squares slope is 1.5. The fitted intercept is b₀ = 11/3 − 1.5(2) = 2/3. |
Determine the least-squares slope for the data points (2, 3), (4, 7), and (6, 11). |
| The least-squares slope is 2, and the fitted straight line is y = 2x − 1. |
Real-world applications
Calibration curves
Experimental physics data analysis
Scientific trend estimation
Engineering regression models
Determining relationships between measured variables
Formula Meaning
The least-squares straight-line slope is the coefficient that minimizes the sum of squared vertical residuals between observed data points and a fitted linear model. It measures the estimated change in the dependent variable for each unit change in the independent variable.
Sources and further reading
No source has been supplied for this item yet. Treat the claim as needing editorial verification.