EconometricsLesson 1 of 1 in the course Ordinary least squares

Multicollinearity: Identification, Diagnostics, and Modeling Responses

5 flashcards · 5 quiz questions · 5 match pairs · 4 fill-in sentences · English

Save course
Lesson progress
0 / 5

Your progress is kept in this browser only.Sign up to save it

Summary

Multicollinearity occurs when predictors in a regression model are linearly related. This lesson introduces the concept and explains why an exact relationship among predictors prevents ordinary least squares (OLS) from identifying unique coefficient estimates.

A regression model uses predictors to explain or predict an outcome. When two or more predictors move together in a way that makes one predictable from the others, their individual contributions can be difficult—or, in the perfect case, impossible—to distinguish.

All flashcards

Perfect multicollinearity
An exact linear relationship makes at least one predictor a linear combination of other model columns.
Imperfect multicollinearity
Predictors have a strong but not exact linear relationship, so separate coefficient estimates may be imprecise.
Variance inflation factor (VIF)
A diagnostic indicating how predictor relationships inflate the variance of a coefficient estimate.
Orthogonal polynomial basis
A representation of polynomial terms using mutually orthogonal columns to improve numerical behavior.
Post hoc analysis
Choosing variables or analysis procedures after inspecting results, potentially undermining conventional inference.

Quiz questions

  1. How should a high variance inflation factor (VIF) for a regression predictor be interpreted?Answer: As evidence of greater collinearity-related uncertainty in its coefficient estimate.
  2. Why can a regression with highly correlated predictors still predict well?Answer: The predictors may jointly carry useful outcome information despite imprecise separate effects.
  3. What can regularized regression methods do when predictors are strongly related?Answer: They can stabilize estimation by adding constraints or penalties.
  4. Why can deleting a predictor solely because of a high VIF be risky?Answer: The predictor may be an important confounder, and deletion can distort estimates.
  5. What is a benefit of an orthogonal change of basis for a regression model?Answer: It can improve numerical behavior while retaining the model's underlying information.

Based on the Wikipedia article Multicollinearity, licensed CC BY-SA 4.0. Adapted into a summary, flashcards and quizzes.

More in Econometrics

All Econometrics lessons
EconometricsCorrelation and Simple Linear Regression in R5 flashcards · 5 quiz questions · 4 gamesPlay free →EconometricsRegression Predictions, Interpolation, and Prediction Intervals5 flashcards · 5 quiz questions · 4 gamesPlay free →