| Join Now
Thank you very much for the reply.
Because of multicollinearity: https://en.wikipedia.org/wiki/Multicollinearity#Consequences
It's pretty well known and commonly taught in stats undegrad so you can find more information by googling multicollinearity.
anyone can help? I would like to get an explanation of the statistical problem below.
suppose we have variables x1, x2 and y. all three are positively correlated. correlation x1 and y is stronger than correlation x2 and y. correlation x1 and x2 is also stronger than correlation x2 and y.
if I regress x2 to y, the result will be positive and significant.
but if I regress x1 and x2 to y, the coefficient x1 must be positively significant. but the x2 coefficient can be significant positive, can be insignificant and can also be significant negative. depending on the difference in the correlation of the three variables. how could this happen?
Statistical Society of Australia
PO Box 213
Belconnen ACT 2616 Australia
02 6251 3647www.statsoc.org.auABN 82 853 491 081
Please direct enquiries to:
Marie-Louise Rankin, Executive Officer
© 2019 Statistical Society of Australia. All Rights Reserved. | website login
Website by Converge Design