STA 302H1F / 1001HF --
Methods of Data Analysis I
-- Fall 2011
What we are covering / what you can skip from the textbook:
-
Chapter 2:
Responsible for all of the chapter.
-
Chapter 3:
You can skip the material on inverse response plots and the Box-Cox procedure in Section 3.3.3.
-
Chapter 4:
You are not responsible for Sections 4.1.1 (PIs for weighted least squares), 4.1.2 (Leverage for WLS), and 4.1.3 (Using LS to calculate WLS).
-
Chapter 5:
Responsible for all of the chapter.
-
Chapter 6:
We are not covering Marginal Model Plots, Inverse Response Plots, and Box-Cox
transformations.
In 6.2.1, the basic idea is important (transform Y and/or X to make the relationship
linear). But we'll just consider one of our standard transformations
(e.g., square root, log, reciprocal, polynomial).
Skip section 6.3.
In 6.4, the text uses some of these tools we aren't covering to find the appropriate
transformation. When reading this section, imagine that the transformations
were decided on as you did in assignment 2.
Ignore the brief bit on marginal model plots in sections 6.4 and 6.5.
-
Chapter 7:
Chapter 7 is about variable selection. If you read it, much of it will be familiar from doing assignment 3, but most of the criteria used will be unfamiliar.
You are only responsible for understanding what you did in assignment 3. The take home message of the assignment is summarized in Section 7.2.3, which you should read.
You may not want to read the rest of Chapter 7, except for
the flowchart on page 252 which outlines the multiple regression
process. The only part of this flowchart which will be unfamiliar is the suggestion that, if everything else is OK but the errors are not normally distributed,
you can use the bootstrap for inference. The bootstrap uses resampling of
the data to numerically approximate the sampling distribution of the
test statistic under the null hypothesis (rather than using theoretical
results, which we did assuming normally distributed errors). You are not
responsible for the bootstrap.