avoir un fort coefficient de corrélation

CONT: L'utilisation de la corrélation ou de la régression [...] informe [...] du comportement de deux ou plus variables entre elles. [...] Seront retenues les variables qui montrent un lien suffisant avec la variable dépendante. (Par exemple celles qui ont un fort coefficient de corrélation ou une signification statistique convenable.)
OBS: Dans les analyses en composantes principales on parle des «coordonnées», du «poids» ou des «coefficients» des variables pour rendre la notion de «loading». Voir la fiche «component loading» pour obtenir plus de renseignements sur cette notion.

CONT: Factor scores ... need to be weighted. Otherwise, the variables which have high loadings on less important factors will have the same impact in the cluster analysis as variables which load highly on important factors.
CONT: The PCA [principal component analysis] was applied to the 240 weather variables for all days within the winter season from 1958/59 to 2000/2001 for 15 Canadian and 12 U.S. stations, producing an 18-component solution that explains 91% of the total variance in the original dataset. The remainder of the components with eigenvalues less than one were discarded. The thermal and moisture variables, air temperature and dew point temperature, load highly on Component 1, explaining over 36% of the total variance.
OBS: See also the record for ""component loading"" for more information on the concept of statistical loading.