16.10 Quadratic Models

In a single linear regression we can include the squared term of the independent variable, that is \(x^2\). In that case, the model is call second-order model. The quadratic term \(x^2\) enable us to hypothesize curvature in the plot of response model relating \(y\) to \(x\).

A quadratic (second-order) model in a single quantitative independent variable is

\[ y_i = \beta_0 + \beta_1 x_i + \beta_2 x_i^2 +\epsilon_i\]

where

  • \(\beta_0\) is the (\(y\)-)intercept of the curve.
  • \(\beta_1\) is a shift parameter.
  • \(\beta_2\) is the rate of curvature.

When \(\beta_2\) is positive, the curve opens upward. When \(\beta_2\) is negative, the curve opens downward.

16.10.1 Example

The data shows the number of weeks employed and the number of errors made per day for a sample of assembly line workers. Find a 2nd order model, conduct the global F–test, and test if \(\beta_2 \neq 0\). Use \(\alpha = 0.05\) for all tests

Errors (\(y\)) and Weeks (\(x\))
errors 20 18 16 10 8 4 3 1 2 1 0 1
weeks 1 1 2 4 4 5 6 8 10 11 12 12
Source: cStatistics for Business and Economics, Global Edition. Chapter 12

The model to be fitted is \(\text{Errors}_i= \beta_0 +\beta_1\text{Weeks}_i + \beta_2 \text{Weeks}_i^2 + \epsilon_i\)

Observations 12
Dependent variable errors
Type OLS linear regression
F(2,9) 174.062
0.975
Adj. R² 0.969
Est. 2.5% 97.5% t val. p
(Intercept) 23.728 21.205 26.252 21.269 0.000
weeks -4.784 -5.747 -3.821 -11.237 0.000
weeks2 0.242 0.171 0.313 7.715 0.000
Standard errors: OLS

The fitted model is

\(\widehat{\text{Errors}}_i= 23.728 - 4.784\text{Weeks}_i + 0.242 \text{Weeks}_i^2\)

  • Considering \(F(2,9) = 174.062\) with $ p-value = 0.000$, \(F\)-test indicates at least one parameter is different from zero.
  • Considering \(t(\beta_2) = 7.715\) with $ p-value = 0.000$, \(t\)-test for \(\beta_2\) indicates indicates curvilinear relationship exists.