# Anova Problem

Discussion in 'P1.T2. Quantitative Methods (20%)' started by sipanivishal, Oct 30, 2008.

1. ### sipanivishalManager-Corporate Banking

Hi David,

http://www.bionicturtle.com/learn/article/anova_table_practice_question_par_4_difficulty/
I just looked at this problem and actually am amazed at the problem itself.F stats is used in multiple regression for testing whether both B2 and B3 are jointly 0 or not.The question appeared to be for bivariate.With df for ESS = 1 no of independent variable is two.Can F test be used for that too.....to check jointly if B1 and B2 are 0 or not.If this is the case then in multiple regression case what does a F test do.....CHecking jointly B1 B2 B3 = 0 or just B2 B3 = 0.

Thanks
Sipani

2. ### David Harper CFA FRMDavid Harper CFA FRM (test)Staff Member

Hi Sipani,

Yes, I realize the F is reviewed in multivariate regression; I don't see a problem with applying it to the bivariate case. Here, the null is simply H0: R^2 = 0.

The F variable stills works: [ESS/1 df]/[RSS/(n-2)].

"Can F test be used for that too.....to check jointly if B1 and B2 are 0 or not"
Yes, that's is main job, to check JOINTLY if the PARAMETERS (not the variables!) are significant. Yes, that would be my two word summary of F: "joint param" check. My question here merely treats the bivariate as a special case of the multivariate. I don't think it's a problem...

David

3. ### sipanivishalManager-Corporate Banking

Hi David,

Surprised as if a new experience....and obviously new out of box questions are very good.Thanks for this wonderful explanation

Thanks
SIpani

4. ### David Harper CFA FRMDavid Harper CFA FRM (test)Staff Member

great, thanks, i will return to your prior question(s) but i need to spend all long day on the cram sessions, so asap on those. thanks, David