Linear Regression 1

Question 1

Plots

Answer 1

• Tukey Anscombe Plot -> linear
• QQ-Plot -> NV
• Scale-Location-Plot -> Homoscedasticity
• Histograms/scatterplots/leverage plot -> find outliers

In R:
Plot()
Autoplot()

Question 2

What to do if assumptions not given

Answer 2

• transformation (log, sqrt)
• take care of outliers
  (Check correctness, transform, chance, interest, only delete with care and mention in report)
• improve model (add interaction)
• use other model family

Question 3

CI

Answer 3

In R:
Confint(lm, level = c(0.95))

Beta1 -/+ c(1.96)*sd

Question 4

Which other regression lines are compatible with observed data?

Answer 4

Confidence band

Question 5

Where do future observations with given x lie?

Answer 5

Prediction band (breiter als confidence band)

Question 6

Make dataset with only few picked variables

Answer 6

Select()

Question 7

Make new data

Answer 7

Expand_grid
Predict
Cbind

Question 8

Get new values of response (y)

Answer 8

Predict()