Understanding Linear Regression Model Properties

A linear regression model describes the relationship between a dependent variable and one or more independent (explanatory) variables. The key characteristic defining a model as "linear" relates to how the parameters (coefficients) enter the equation.

Linearity in Parameters

A regression model is considered linear if the equation is linear in its parameters. This means the parameters are simply added or subtracted, possibly multiplied by variables or functions of variables, but they are not exponents, roots, or inside trigonometric, logarithmic, or other non-linear functions.

For example, the model:

$ Y = \beta_0 + \beta_1 X_1 + \beta_2 X_2 + \epsilon $

is linear in the parameters $\beta_0$, $\beta_1$, and $\beta_2$. Here, $Y$ is the dependent variable, $X_1$ and $X_2$ are explanatory variables, and $\epsilon$ is the error term.

Linearity in Variables

The relationship between the dependent variable and the explanatory variables does not strictly have to be linear. A linear regression model framework can accommodate non-linear relationships with the variables by transforming the variables themselves.

For instance, consider the model:

$ Y = \beta_0 + \beta_1 X + \beta_2 X^2 + \epsilon $

This model is still considered a linear regression model because it is linear in the parameters $\beta_0$, $\beta_1$, and $\beta_2$. Even though the relationship between $Y$ and $X$ is quadratic (non-linear), we can treat $X^2$ as a new variable (say, $X_{new} = X^2$). The model then becomes $Y = \beta_0 + \beta_1 X + \beta_2 X_{new} + \epsilon$, which is linear in the parameters and involves the variables $X$ and $X_{new}$.

Evaluating the Options

• Option 1: Incorrect. Linear regression models must be linear in parameters.
• Option 2: Incorrect. Linear regression models must be linear in parameters, not non-linear.
• Option 3: Incorrect. While it must be linear in parameters, it does not *have* to be linear in variables; transformations are allowed.
• Option 4: Correct. This accurately reflects that the model must be linear in its parameters ($\beta$'s), but the variables ($X$'s) can be transformed, allowing for non-linear relationships between the dependent and independent variables (e.g., polynomial regression).