Week 10 Summary (DSME5110F)
Multiple Linear Regression
• Select the subset of predictors – Why?
∗ Need more data ∗ Overfitting
Copyright By PowCoder代写 加微信 powcoder
∗ Multicollinearity
∗ Exhaustive Search (regsubsets() in package leaps) ∗ Automatic stepwise regression techniques
· forward, backward, stepwise (step() with option direction = “forward”, “backward”, or “both”)
• Dummy variables
– If a categorical variable has k categories, then we will need k − 1 dummy variables. – By default, R uses the first level of the factor variable as the reference.
• Extensions
– Non-linearity (define new variables; or use I()), interaction effect (x1 ∗ x2) – Log Transformation (log())
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com