This is a session of Topic 3: Statistics education at the post-secondary level Full topic list
(Monday 3rd, 14:00-15:30)
Teaching robust methods
Most classical statistical procedures are derived under the assumption that all the observations follow a single statistical model, usually with normal errors. These procedures have optimal statistical properties under this model. However, they are very sensitive to departures from these assumptions. Small deviations from normality or a few outliers — that is, atypical observations that do not follow the model — in the sample may have a very large influence on the results of classical methods and lead to erroneous conclusions. A statistical procedure is called robust if (i) It is resistant to the presence of outliers and another model violations, (ii) It is highly efficient under the assumed model. The speakers in this session will address the following topics:
- Motivation of robust procedure via real data applications.
- Historical perspective of the development of Robust Statistics.
- The teaching of robustness, including the use of interactive graphics.