Statistics 621

First Quarter, 2001

Dr. Robert Stine


Click here to get to the class web page which is common for all sections.

Class lecture notes

This outline indicates the topics and data sets that will be covered in class.

  1. Fitting equations to data

  2. Assumptions in regression (9 Sep 2001)

  3. Prediction and confidence intervals (11 Sep 2001, small edits)
    We will use this JMP-IN script to explore how outliers affect a regression model. For additional discussion of logs in regression, I have an extra handout with more examples.

  4. Multiple regression (16 Sep 2001)

  5. More multiple regression (17 Sep 2001)
    For some extra examples on interpreting multiple regression, we will use these handouts (I will distribute a copy in class): The data sets for these examples are We will then continue with the car data.

  6. Collinearity in regression (20 Sep 2001)
    This class completes our discussion of collinearity in multiple regression, focusing on diagnostics and possible remedies. Alas, a scheduled fire drill will interrupt our coverage of this material and hold things back. For the assignment, be sure to have a look at the discussion of the partial F test in the casebook, pages 151-152 in the parcel handling example. The illustration of partial F in the regression for fuel consumption (car89) is complicated by the presence of missing data for some predictors.

  7. Categorical predictors in regression (revised, 24 Sep 2001)

  8. Categorical predictors (26 Sep 2001)

  9. Categorical terms with many levels (revised, 2 Oct 2001)
    We will also use this FedEx example to review material from the previous class. It illustrates a two-group example with an interaction. Time permitting, we will begin our discussion of the model-building process.

  10. Building regression models (3 Oct 2001)
    We will conclude our analysis of the use of categorical factors in multiple regression with a quick look at a topic discussed back in Stat 603 - namely the issue of multiple comparisons . Most of our time will be spent on the modeling process.

  11. Diagnostics for regression models (9 Oct 2001)
    This lecture concludes our discussion of the project data and multiple regression models. We will discuss various types of residual diagnostics, such as those that indicate that you have left out a factor from the model. We will also take a look at some sample executive summaries from last year's project.

  12. Analysis of variance (preliminary, 10 Oct 2001)
    This lecture concludes the course with a look at a different approach to regression, one based on data gathered in highly structured experiments. All of the predictors are categorical, and interaction becomes yet more interesting and revealing. The methods are closely related to conjoint analysis as used in marketing. The two examples from the casebook illustrate the analysis both with and without an interaction.

Data sets from the casebooks

You can get the data for the casebook examples either as a single zip file or as individual JMP files from this link. If you take this route, note that the assignments in this zip file are for those in the casebook, not those that we are using this term.

Other interesting links

Here are some other links related to the course that you might find interesting...

Here are other links of general interest...

Practice questions

In addition to the "continuous assessment" questions on the class web page, you may find these useful to check your understanding of the material.

Supplemental Handouts

Screen capture video clips