Multiple Regression

Many researchers have to analyse data with the aim of finding an equation that relates a dependent y-variable to one or possibly more explanatory x-variables. This procedure is generally known as Multiple Regression. While many computer packages will perform the necessary number crunching, they still assume that you know the underlying theory and can interpret their output.

The Statistical Consulting Service will run a One-Day Course on multiple regression.

Content

The course will deal with the problem of finding the equation that 'explains' a y-variable in terms of one or more x-variables, where the y-variable and the x-variables contain real numbers (for example, the y are not labels for success/failure, or gender, or ethnic background). The course will not cover situations where any of the variables are categorical, nor generalized linear models, nor other topics that are extensions of multiple regression.

The material will consist of

Overview
Simple Linear Regression (one y-variable, one x-variable)
Fitting a straight line to the data
What's the computer doing?
How do I get the computer to do the right thing?
What does the output mean?
How do I check that the statistical assumptions are valid? What if they are not valid?
What if my data points are not equally important?
What if some points have more influence on the equation than others?
Multiple Regression (one y-variable, several x-variables)
(a) Fitting a simple plane to the data. We answer the set of questions listed in Section.
(b) Fitting a curved surface to the data. Why would I want to fit a curved surface? We answer the set of questions listed in Section 2.
How can this material be extended?

For further information about course content, contact Chandra Gulati (chandra_gulati@uow.edu.au).

Enrolment

A charge of $66 (or $60 if paid by your department) per person will be made, and this includes the cost of materials. A receipt will be issued. The number of available places is limited. Enrolments will be taken strictly in order of receipt. Details of the venue of the course will be provided after enrolment. The course may not run if there are insufficient enrolments.

Assumed Knowledge

Some prior statistical knowledge, such as the concepts of null and
alternative hypotheses, p-values and confidence intervals. Having completed "Statistical Methods for Research Workers" will be very helpful.

Statistical Package Used

The course will use the package SAS, for which the University has a site licence. Full instruction in using SAS for multiple regression will be given. Attendees who are proficient with another package can use SAS for the day, and then subsequently apply the techniques to their package of choice by following the notes that will be issued. All attendees will have the individual use of a computer at the course. For those without experience with SAS, attendance at the "Introduction to SAS" course will be very helpful.

Last reviewed: 4 June, 2009