Hello everyone
Could anyone through some light on this...
Can multicollinearty still exist in a multiple regression model produced by
stepwise variable selection (where each of the predictors in the model are
significant)?
I have read that multicollinearity can produce coefficients of X which
appear to be non significant when there is actually a significant
association between X and Y in the population.....the variances of model
coefficients and also inflated. Bearing this in mind I would like to know
if *significant* coefficients in a model can still be adversely affected by
multicollinearity.
Many thanks in advance,
Kim.
|