In the first part of the paper, we will develop the theory and practice of regression modelling, which is the name given to developing statistical models which predict a variable y from one or more other variables. In STATS221, you will have met this situation in the case of linear regression where y is assumed to be normally distributed about a mean given by a linear combination of other variables. We will review linear regression using the R software and then go on to show models can be developed where the distribution of y about its modelled average value is non-normal. These are known as Generalised Linear Models (GLMs). The use of these models to analyse appropriate data sets will be taught.

In the second part of the paper, we will introduce Weka and study several topics in the area of Multivariate Statistics and Machine Learning: Classification and Discrimination; Regression using Weka; Tree Models; Principal component analysis, Multivariate distances; Unsupervised learning – hierarchical and non hierarchical; the K means clustering algorithm; and the EM algorithm.

The paper is broadly divided into two main topics:

• Generalised Linear Models, taught in Weeks 1 to 6; and
• Multivariate techniques [jointly with the students of COMP321], taught in Weeks 7 to 12.

Instruction is given in the use of the free, open-source statistical software R and Weka.

During the first six weeks, students of STAT321 will attend three lectures and one tutorial per week. During the second six weeks, STAT321 students will join COMP321 students in the COMP321 stream, attending two lectures and one lab per week. Note that the times and locations of lectures will change from the first half of the course to the second.

Each half of the paper contributes half of the internal assessment.

The internal assessment will consist of:

Weeks 1-6 assessment:

2 assignments (7.5% each) and one test (10%).

Weeks 7-12 assessment:

4 tutorial exercises (3.75% each) and one assignment (10%).

Note: The internal assessment component contributes 50% to your final mark, and the final exam contributes the other 50%.

A final mark of 50% or higher is required to pass the course. However, note that if you manage to achieve more than 50% overall, but less than 20% of your final mark is from the exam, you will receive a restricted pass.

All material provided for this paper can be accessed from Moodle. Moodle will also be the source of news concerning the paper. It is your responsibility to ensure that you have access to the STAT321 page on Moodle, and your contact details on Moodle are up-to-date. To access Moodle, go to http://elearn.waikato.ac.nz.

Students should expect to spend a minimum of about 10 hours per week on this paper. This includes the 4 scheduled contact hours involving lectures, tutorials, and labs.