above 10 is fairly high. MC does not effect model estimates but can blow out their SE. there are options such as dropping one of the variables.
Hi guys, im relatively new in this forum. I have a question, i am using multiple regression analysis for my thesis. The result shows the P-value is less than 0.05 (regression model significant) but the VIF value of some depend factor is over 10. Is is Multicollinearity? Will i be able to use this model to predict independent? Is there any reference?
above 10 is fairly high. MC does not effect model estimates but can blow out their SE. there are options such as dropping one of the variables.
Stop cowardice, ban guns!
if the p-value is still below 0.05 ths should not be a worry, I believe.
regards
the vif of some dependent var over 10, which one? tell more.
Stop cowardice, ban guns!
In the first regression model, I have 6 factor and all of them are over 0.05, there is only 2 factors with VIF less than 10. F test shows that its not significant
In the second regression model, i have 6 factors and all of them are less than 0.05, there is only 3 factors with VIF less than 10. F test shows that its significant.
I have not done tolerance test fyi.
In this case it might happen that some of the factors are significant but you miss it because the confidence intervals are inflated, so the large VIF could be a problem.
If the factors are significant even though the CFs are inflated you do not need to worry about the VIFs.
regards
LizBeth (05-30-2016)
Thanks Rogojel for the fast reply, i think the same way too . Could you recommend be some journals or book to backup the statement. That would surely help alot.
Hi,
I just used simple reasoning about the confidence intervals but now that I googled it this seems to be really good. Look at the comments as well. http://statisticalhorizons.com/multicollinearity
LizBeth (06-01-2016)
Thank you so much rogojel, that has been very helpful for my thesis )
Tweet |