Multiple linear regression model violations

25 important questions on Multiple linear regression model violations

Q: What are the four main MLR model violations (Ch. 22 Nieuwenhuis)?

  • Collinearity
  • Heteroskedasticity
  • Non-linearity & non-normality
  • Dependence (autocorrelation) of error terms

Q: What is collinearity?

An independent variable is strongly related to a linear combination of other independent variables.

Q: What is heteroskedasticity?

The variance of error terms is not constant:
Var(εi) ≠ σ².
  • Higher grades + faster learning
  • Never study anything twice
  • 100% sure, 100% understanding
Discover Study Smart

Q: What is non-linearity in regression?

The linear model does not match reality: E(Y | X) is not linear in X.

Q: What is non-normality of errors?

Residuals are not normally distributed → assumption violated.

Q: What is autocorrelation?

Errors are correlated; typically εₜ and εₜ₊₁ move together → common in time series.

Q: Why is autocorrelation a problem?

  1. LS estimators not optimal
  2. t-tests and F-tests invalid
  3. Variances underestimated
  4. Results not trustworthy

Q: What does the Durbin–Watson test check?

Whether first-order autocorrelation exists in the errors.

Q: What is the range of DW?

0 to 4

Q: How to interpret DW values?

  • D ≈ 2 → no autocorrelation
  • D → 0 → positive autocorrelation
  • D → 4 → negative autocorrelation
  • Q: Positive autocorrelation test H₁: positive autocorr → when do we reject H₀?

    Reject H₀ if D ≤ dL

    Q: Negative autocorrelation test H₁: negative autocorr → when do we reject H₀?

    Reject H₀ if D ≥ 4 − dL

    Q: Two-sided autocorrelation test → when reject H₀?

    Reject H₀ if
    • D ≤ dL (positive)
    • D ≥ 4 − dL (negative)

    Q: What are the three possible conclusions of a DW test?

  • Reject H₀
  • Accept H₀ (no autocorr)
  • Inconclusive (between dL and dU)
  • Q: What is a dummy variable?

    A variable that takes value 1 if a category is present, and 0 otherwise.

    Q: Why do we use dummy variables?

    To include qualitative (categorical) variables in regression.

    Q: What is the “base level”?

    The category represented by all dummies = 0, used as reference group.

    Q: Interpretation of B for a dummy variable?

    The difference in means between the dummy category and the base level, holding other X’s constant.

    Q: How do we interpret β for ‘female’ dummy (1 = female, 0 = male)?

    β = mean(female) – mean(male).
    If β < 0 → women earn less (in wage example).

    Q: What does the partial F-test check?

    Whether a subset of variables is jointly useful in the model.

    Q: Hypotheses for partial F-test?

    H₀: the extra variables are not useful (all β’s = 0).
    H₁: at least one β ≠ 0.

    Q: What is the test statistic for partial F-test?

    Formula:

    Q: What are the degrees of freedom?

    Numerator: k − g
    Denominator: n − (k + 1)

    Q: When do we reject H₀ in a partial F-test?

    Formula:

    Q: What does a huge F value indicate?

    Strong evidence in favour of H₁ → variables jointly useful.

    The question on the page originate from the summary of the following study material:

    • A unique study and practice tool
    • Never study anything twice again
    • Get the grades you hope for
    • 100% sure, 100% understanding
    Remember faster, study better. Scientifically proven.
    Trustpilot Logo