Step back a minute: normality is NOT required for predictors in a
multiple regression model, though the sqrt(x) transformation may
also make the relationship more nearly linear, and linearity IS
assumed when you fit a simple model such as y ~ x + w + z.
(Normality is only required for the residuals/errors)
To see what's going on, you can make make partial regression /
added-variable plots using car::avplots. The loess smooth will
show you if the relationship is non-linear.
HTH
-Michael
Em 23-10-2017 18:54, kende jan via R-help escreveu:
Dear all, I am trying to fit a multiple linear regression model with a
transformed dependant variable (the normality assumption was not
verified...). I have realised a sqrt(variable) transformation... The
results are great, but I don't know how to interprete the beta
coefficients... Is it possible to do another transformation to get
interpretable beta coefficients to express the variations in the
original untransformed dependant variable ? Thank you very much for
your help!Noémie
[[alternative HTML version deleted]]
______________________________________________
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
