Correspondence Analysis

Correspondence Analysis is a factorial analysis technique applied to the study of data tables whose cells contain either frequency values (real positive numbers) or presence-absence values ("1" or "0").

Like all factorial analysis techniques, correspondence analysis allows the extraction of new variables - the factors - with the property of summarizing in a organized way the significant information contained in the countless data tables cells; furthermore, this analysis technique allows the creation of graphs showing - in one or more spaces - the points that detect the objects in rows and columns, that is - in our case - the linguistic entities (words, lemmas, texts segments and texts) with the respective source features.

In geometrical terms, each factor sets up a spatial dimension - that can be represented as an axis line - whose center (or barycentre) is the value "0", and that develops in a bipolar way towards the negative (-) and positive (+) end, so that the objects put on opposite poles are the most different, almost like the "left" wing and the "right" wing on the political axes.

In T-LAB the analysis results are summarized through graphs that allow the evaluation of the relationships of proximity/distance - or rather similarity/dissimilarity - between the considered objects.

Furthermore, T-LAB shows measures (i.e. Absolute Contributions and Test Value) that help to understand the poles of factors that set up similarities/dissimilarities between the considered objects.