Linear Regression Standard Error

Routine Summary

Calculates the standard error associated with linear regression coefficients.

Inputs

L₁ - values of the independent variable
L₂ - values of the dependent variable

Outputs

Ans - a 2-element list containing the standard errors

Variables Used

L₁, L₂,

Calculator Compatibility

TI-83/84/+/SE

:2-Var Stats
:LinReg(ax+b)
:a√((r²ֿ¹−1)/(n-2)){1,√(Σx²/n)}

This routine computes the standard error (uncertainty) associated with the linear regression coefficients a and b (σ_a and σ_b, respectively) for the regression equation y=ax+b. Precisely stated, the true value of the coefficient a is expected to be within the interval a±σ_a, and similarly for b.

The routine returns a two-element list; σ_a is the first element, and σ_b is the second element.

If one prefers to use the function LinReg(a+bx) instead of LinReg(ax+b), the appropriate routine is:

:2-Var Stats
:LinReg(a+bx)
:b√((r²ֿ¹−1)/(n-2)){√(Σx²/n),1}

(note that the meanings of σ_a and σ_b have now interchanged).

In both routines, r², a, b, n, and Σx² are statistical variables.

Formulas

For the fitting equation y=ax+b,

(1)

\begin{align} \definecolor{darkgreen}{rgb}{0.90,0.91,0.859}\pagecolor{darkgreen} \begin{align*} \sigma_a&=a\sqrt{\frac{\frac1{r^2}-1}{n-2}} \\ \sigma_b&=\sigma_a\sqrt{\frac{\Sigma x^2}{n}} \end{align*} \end{align}

where n is the number of data points, r² is the coefficient of determination, and Σx² is the sum of squares of the independent variable values.

Error Conditions

• ERR:DIM MISMATCH is thrown if the two lists' sizes are not the same.

Reference

Lichten, William. Data and Error Analysis., 2nd. ed., Prentice Hall: Upper Saddle River, NJ, 1999.

.