MH4501 Multivariate Analysis
Course Summary
The course looks into methods to understand higher dimensional data. The first half of the course looks into multivariate statistics and hypothesis testing, while the second half deals with more machine learning-ish concepts. This course has a lot of linear algebra so make sure your linear algebra foundations is strong. The topics covered are:
- Matrix algebra
- Population and sample statistics
- Multivariate normal distribution (and related multivariate distributions)
- Multivariate statistical inference
- Multivariate analysis of variance (MANOVA)
- Principal component analysis (PCA)
- Factor analysis (FA)
- Discriminant analysis (DA)
- Clustering analysis (CA)
- Canonical correlation analysis (CCA)
Workload and assessment
The content itself is not that difficult, but the lectures will have significant amount of proving of theorems and results (which can be tricky). The workload is moderate: there are 4 graded assignments, 1 midterm exam and 1 final exam. There may be a lot of matrix calculations done by hand especially during exams which can be pretty tedious, so having a graphic calculator will come in handy.
Conclusion
The course seems to be a useful course for students. In the first half, students learn how to extend their current knowledge of statistical inference to a multivariate perspective, which is useful for data analytics. Furthermore, the second half is useful so that students do not merely use machine learning algorithms as a black box, and instead understand the mathematical principles behind the algorithms.