Multivariate exploratory techniques for linguists with R
I taught a four-hour online workshop on multivariate exploratory techniques on Monday, July 13, 2020. This event was part of the 14th Language and Teaching Corpora conference, which was held at the University of Perpignan (France).
What the workshop was about
Once corpus linguists have collected sizeable amounts of observations and described each observation with relevant variables, they look for patterns in the data. When the data set is too large, it becomes impossible to summarize the table with the naked eye, and summary statistics are needed. This is where exploratory data analysis steps in.
I introduced five multivariate exploratory techniques:
- correspondence analysis,
- multiple correspondence analysis,
- principal component analysis,
- exploratory factor analysis,
- multidimensional scaling.
These techniques rely on an attempt to simplify complex multivariate datasets to facilitate interpretation.
Each method was illustrated with case-studies related to the themes of the TaLC2020 conference.
Participants learnt how to:
- collect and/or load corpus data into R;
- detect and avoid the common pitfalls of quantitative analysis;
- run multivariate exploratory statistics;
- interpret and report the results in a research paper.
Although no expert knowledge of R was required, a basic knowledge of computers and office software was strongly recommended.
The notebook that was used during the workshop can be accessed via the following link: