PROGRAM

Since the type of data analysed by the target audience for this course involves a very high number of compositional parts, the topics to be treated will include the following:

1. The use of classical SVD-based methods of dimension reduction, called logratio
analysis in this context

 

2. Clustering of parts as a way to reduce dimensionality

 

3. Variable selection as a way to reduce dimensionality

 

4. The problem of zeros in compositional data

 

5. The correspondence analysis alternative, which can be measurably close to the logratio approach, and which has no problem with data zeros.

 

The R programming environment will generally be used in the practical part of the workshop. Participants should preferably have their own laptops, with R and RStudio installed.