multicollinearity CFAI topic test doubt

D

Deal_Clincher

New member
Joined
Feb 28, 2015
Messages
0
Reaction score
0
According to CFAI Option B is correct.
So, if just one of the coeffs negates the H0 in t-test there is no multicollinearity. Is that right?
Source : Quantitative Methods - Garfield - CFAI Sectional Test
 
The CFA Institute doesn’t do a very thorough job with the topic of multicollinearity. In this case, only two independent variables are used, so you could look at a pairwise correlation coefficient between the variables (although I don’t see one given). For this example, it seems they want you to see that the f-test is significant, r-squared is reasonably “high”, and that at least one of the slope coefficients is significant (not how they describe classic indications of MC). So for this problem, it seems that will work.
In the real world, and approaching this more correctly, a significant t-test doesn’t guarantee multicollinearity isn’t an issue (you can have significance due to a large sample size but still have multicollinearity cause issues with estimated coefficients, for example).
You might also want to remove protected intellectual property from your post.
 
tickersu wrote:
The CFA Institute doesn’t do a very thorough job with the topic of multicollinearity. In this case, only two independent variables are used, so you could look at a pairwise correlation coefficient between the variables (although I don’t see one given). For this example, it seems they want you to see that the f-test is significant, r-squared is reasonably “high”, and that at least one of the slope coefficients is significant (not how they describe classic indications of MC). So for this problem, it seems that will work.
In the real world, and approaching this more correctly, a significant t-test doesn’t guarantee multicollinearity isn’t an issue (you can have significance due to a large sample size but still have multicollinearity cause issues with the practical meaning of estimated coefficients, for example).
You might also want to remove protected intellectual property from your post.
Click to expand...
Thank you
 
You must log in or register to reply here.
Back
Top