Applied statistics (ESB: RISK)

Frontpage Lecture plan Vidcasts Datasets R install help

Multiple linear regression

Literature

[WMMY] Chapter 12.1, 12.2, 12.4, 12.5 until p.476m, 12.6, 12.8

Lecture material

This lecture as: slideshow (html), Rmarkdown (Rmd), notes (pdf).

The entire module: notes (pdf).

Exercises

  1. Consider the following prediction equation without interaction:  = 2 + 0.4x1 + 0.7x2. What does the equation for the predicted response look like when x1 = 1? What does the equation for the predicted response look like when x1 = 5?

  2. Consider the following prediction equation with interaction:  = 2 + 0.4x1 + 0.7x2 − 0.1x1x2. What does the equation for the predicted response look like when x1 = 1? What does the equation for the predicted response look like when x1 = 5? Explain the difference between the models in 1. and 2.

  3. For this exercise we use data from [WMMY] Exercise 12.5. You find the exercise in the Rmarkdown file Exercise_12-5.Rmd.

  4. Answer the questions below using the Rmarkdown file GNP.Rmd.

    • In this exercise we consider a dataset containing macro economical numbers for USA collected in the years 1947 to 1962. The dataset contains 7 variables:
      1. GNP.deflator: GNP implicit price deflator
      2. GNP: Gross National Product
      3. Unemployed: Number of people that are unemployed
      4. Armed.Forces: Number of staff in the armed forces
      5. Population: Population size (age >=14)
      6. Year: Year
      7. Employed: Number of people that are employed
    • Make a multiple linear regression model with ‘GNP’ as response and ‘Population’ as explanatory variable.
    • What is the interpretation of the estimates? Is the population size significant for GNP?
    • Repeat the analysis above, but now with ‘Year’ as explanatory variable. Is time significant?
    • Repeat the analysis above, but now with both ‘Population’ and ‘Year’ as explanatory variables. Conclusion?
    • Make pairwise scatter plots (e.g. ggscatmat from the GGally package) of GNP, Population and Year
      • Does the plot explain your results?