PDA

View Full Version : T- statistic derivation on page 151 of econometric models

Gareth Keenan
07-13-2006, 01:37 AM
Would someone throw me a bone and let me know how to derive the stinking T-statistic 4.4 and 15.9? Preferably with out just taking the square root of the F test statistic(which I was successfully able to derive on my own) I'm missing something, and am perfectly cool with bringing shame on House Keenan for revealing I don't know something I should. Could some just throw me a bone.

And, I think that of Exams 1,2, CAS 3, and NEAS online regression analysis, that page for page the regression analysis is the most conceptually challening. There...I said it.

Attached should be the spread sheet that I was trying to work with. It's not very pretty, but it's explanatory. It's actually an .xls, just rename the extension when you download it(if you feel like it)

OK, I'm going home now, it's 10:38, and I've been studying at work since 5:30

Gareth Keenan

4199

bmathew22
07-13-2006, 07:59 AM
(estimator-null hypothesis value for estimator)/standard error for estimator

Gareth Keenan
07-13-2006, 12:19 PM
What's my null hypothesis, that alpha and beta equal zero? Because that I can do.

louisdj_KSU
07-13-2006, 12:38 PM
Use equation (3.18) for beta (and change the betas to alphas) on pg. 68. For s values, see the gray box on page 65.

I was unable to open your file (some messeage saying the file didn't START with .pdf?!?). Hence I'm writing this with no clue of what you've done. If you want to put some of the numbers you got (like the s's, ESS, TSS) go ahead and post them. I'll tell you if I got that too.

louisdj_KSU
07-13-2006, 12:38 PM
What's my null hypothesis, that alpha and beta equal zero? Because that I can do.

YES!

Gareth Keenan
07-13-2006, 09:17 PM
I Got it to work, just Beta Hat over Beta standard deviation for the Beta T-statistic. Thanks

Still...Why do we choose the null hypothesis to have Beta be equal to zero? Is it just because it makes the calculation easier, or is the idea more along of the lines of...We don't need this coefficient at all to model the dependent variable, then see what your statistic is and then it give you a better idea if you need it or not.

Cheers,

Gareth Keenan.

louisdj_KSU
07-13-2006, 10:10 PM
I Got it to work, just Beta Hat over Beta standard deviation for the Beta T-statistic. Thanks

Still...Why do we choose the null hypothesis to have Beta be equal to zero? Is it just because it makes the calculation easier, or is the idea more along of the lines of...We don't need this coefficient at all to model the dependent variable, then see what your statistic is and then it give you a better idea if you need it or not.

Cheers,

Gareth Keenan.

In short, I think it's answering the question "Does Y really depend on the X varibable associated with this beta (hat)?" Another way of asking is "Should the list of independant variables for our model include this X or not?"