Hello S-users,
how should one interpret correlations amongst regression coefficients as
outputted by summary.glm when some, or all, the regressors are categorical
variables coded with the treatment contrast? Do those correlations mean
something? If so how are they calculated? Should one worry when the
reported correlations for some levels of the categorical variables are
high?
Obviously the levels of a single categorical variable will be highly
"related" amongst themselves, even after removing the reference level's
column, does this have a negative impact on the estimation of the
coefficients?
Thank you very much for your insights,
Gérald Jean
Analyste-conseil (statistiques), Actuariat
télephone : (418) 835-4900 poste (7639)
télecopieur : (418) 835-6657
courrier électronique: gerald.jean@spgdag.ca
"In God we trust all others must bring data" W. Edwards Deming
|