Correspondance Analysis

Correspondance analysis provides original visualizations of factor analysis as computed using FactoMiner library in R.

The parameters are quite intuitive. First one has to decide on the nature of individuals that will be considered to build the Burt matrix. For a bibliometrics analysis, one could work at the level of individuals, institutions or papers for instance.

Then the active variables should be entered (as many dimension as needed). Finally the last option allows to also include supplementary variables that will not contribute to the definition of the axes.

Three interactive visualizations are released. they are all bi-dimensional and show the positioning of variable modalities along the three first dimensions. The first html file shows dimensions 1 and 2. The second html file shows dimensions 2 and 3. The third html file show dimensions 1 and 3. Finally the  fourth html file codes first and second principal component on the x and y axes and use color for the 3rd component. Other files are produced containing the detailed positions and contributions of each variable to the correspondance analysis.

 

As an illustration, see the result of a correspondance analysis mixing named entities extracted from the bible:

Correspondance Analysis with 3*10 Named Entities (PER + GPE + ORG) and bible books as a supplementary field

learn about CorText scripts and share your experience