Multiple Linear Regression
Multiple Linear Regression
Examines relationship between one dependent variable and multiple independent variables.
Benefits Over Simple Regression
Incorporates multiple predictors for combined effects.
Improved predictive accuracy through simultaneous consideration of predictors.
Controls for confounding variables, isolating unique contributions.
Assesses relative importance of each predictor variable.
Assumptions Pre and Post-Test
Pre-Test: 1) Linear relationship between variables; 2) No multicollinearity (correlation < .80).
Post-Test: Multivariate normality of residuals; homoscedasticity of residuals.
Example Scenario: Real Estate
Dependent Variable: House prices
Independent Variables: Size, number of bedrooms, bathrooms, distance from city centre, violent crime rate.
Regression Statistics Overview
Sample Data: Shows different predictors and their influence on house prices.
Key Assumptions of Regression
Continuous variables in pairs; linear and homoscedastic relationships.
Multiple Regression Model Creation
Focused on factors like house size, number of bedrooms, etc.
Importance of checking for multicollinearity via correlation matrices.
Important Metrics
Adjusted R-squared: Proportion of variance explained by predictors.
Coefficients: Estimated changes in dependent variable per unit change in predictor.
Significance Levels (p-values): Determine statistical significance of coefficients.
Homoscedasticity and Normality
Residuals should show consistent variance (homoscedasticity) and normal distribution.
Key Findings from Model
Significant predictors: Number of bedrooms, distance to city.
Non-significant predictors: House size, number of bathrooms, violent crime rate.
Takeaways
Multiple regression assesses effects of multiple variables on one dependent variable.
Useful for prediction, but does not confirm causality due to possible omitted variables.