The table 7 (that we have divided in two parts so it can be seen better) shows us
which are the segments that allow us to identify better each group (9). The first, in which the identity
is "Spanish / not - Spanish". The second fact that divides the group is the
auto-categorisation as a Catalan speaker / non Catalan speaker", together with the
representation of Catalan. In the third level we will find the variables: Social Catalan
net, representation of Catalan and representation of Spanish. As we see this method allows
us to consider other variables as well as the ones proposed in the hypothesis (the
identity "Spanish / not - Spanish" and the auto-categorisation as a
"Catalan speaker / non Catalan speaker") and obviously we will find them in the
most important places.
In the table number
8, referring to the year 2000, we find once more the same three variables. Only in the
third level, little changes are produced, like the one of the order in which the nets and
representations appear. This fact shows us the stability of the characteristics from the
linguistic groups. It is very interesting to compare the punctuation obtained in both
researches because they show us the evolution of the linguistic groups, but, sadly, we
dont dispose of enough space to comment them know.
4.3.
Predictive results
4.3.1. The predictions of belonging to a linguistic group
The model that we
have exposed allows us to preview the group in which the pupil belongs to without having
the information that has taken us to do this classification (10). The one of 1993 previews
an 83.8% of the classified cases, and in no way it does any mistakes with a group that is
not neighbour. In year 2000, the level of right answers decreases in 72.3%, but it
doesnt do either any mistake that is not with a adjacent group, therefore, these
elevated percentages show us the validity of the model that we have used. Let us see in
tables 9 and 10.
Table 9. Choice
level (83.8%) of the model (1993)
Table 10. Choice
level (72.3%) of the model (2000)
|