Postgraduate - UnitEPM5004 - Linear models

This unit entry is for students who completed this unit in 2014 only. For students planning to study the unit, please refer to the unit indexes in the the current edition of the Handbook. If you have any queries contact the managing faculty for your course or area of study.

print version

6 points, SCA Band 2, 0.125 EFTSL

Refer to the specific census and withdrawal dates for the semester(s) in which this unit is offered, or view unit timetables.

Level	Postgraduate
Faculty	Faculty of Medicine, Nursing and Health Sciences
Organisational Unit	Department of Epidemiology and Preventive Medicine
Offered	Alfred Hospital Second semester 2014 (Off-campus)
Coordinator(s)	Professors A Forbes & J Carlin

Synopsis

This unit explores biostatistical applications of linear models with an emphasis on underlying theoretical and computational issues, practical interpretation and communication of results. By a series of case studies, students explore extensions of methods for group comparisons of means (t-tests and analysis of variance) to adjust for confounding and to assess effect modification/interaction, together with the development of associated inference procedures. Multiple regression strategies and model selection issues will be presented together with model checking and diagnostics. Nonparametric regression techniques, and random effects and variance components models will also be outlined.

Outcomes

On completion of this unit students should be able to:

understand the major theoretical and computational issues underlying analyses based on linear models;
develop appropriate regression modelling strategies based on unit matter considerations, including choice of models, control for confounding and appropriate parameterisation;
be proficient at using a statistical software package (eg. Strata) to perform multiple regression and analysis of variance;
understand the construction, use and interpretation of regression modelling diagnostics;
express the results of statistical analyses of linear models in language suitable for communication to medical investigators or publication in biomedical or epidemiological journal articles; and
appreciate the role of modern techniques including nonparametric smoothing and variance components models.