Packet 6

Packet 6

1. Qualitative Predictors

   • Categorical predictors = Factors
     • Levels
       • For k levels, need k-1 indicator variables

2. Regression with Principal Components

   • 1. Normalize the X's
     • scale(X)
   • 2. Compute Principal Components
     • prcomp(normalizedX)
   • 3. Decide how many PC's to use
   • 4. Fitting the model with PC's
   • Quantitative data only

3. Variable Screening Methods

   • Simplicity
   • Powerful
   • All Possible Regressions Selection Procedure
     • Criterion-based
       • Adjusted R^2
       • AIC
         • stepAIC()
       • R^2

4. Multicollinearity

   • Correlation between predictors
   • Effects
     • Inflated standard errors
     • Decreased test statistics
     • Increased p-values
       • Test not valid
   • Methods of dealing with
     • Variable selection
     • PCA
     • LASSO/Ridge Regression
   • VIF
     • vif()

5. Greta Klaus, Mitchell Larson, Josie Wesseler

Bilder und Dateien der Mindmap