beta=1 test

misi · November 23rd, 2005, 9:42 am

Hi all,I'll go straight to the question...In OLS, do you know I could test the significance difference from 1 (instead of zero) of the slope coefficient?Thanks all.misi

DimitrisLancs · November 23rd, 2005, 11:31 am

You have to use the t-student again. Hypothesis:Ho: beta=1 H1: beta<>1 (instead of 0)Estimate t as:t= (beta - null hypothesis)/std.error(beta)Compare t with the t statistic from the back of your statistics book (with your required significance level)

misi · November 28th, 2005, 2:26 pm

Thanks a lot dimitris!However, I should verify that beta is equal 1 (H1)...not its difference from 1 (H0). Could you help me, please?thanks again.misi

pcsiung · December 17th, 2005, 12:23 am

2 options really,like what dimitris said, t-stat = [beta1 - 1 (Ho)]/SE (beta1), instead of the usual (beta1-0)/SEanother way is to rearrange your model. try to make your model, Y = beta0 + beta1*X + einto:- (Y-X) = beta0 + (beta1 - 1)*X + e, that way you may use the normal t-stat formula.most softwares i think test for beta = 0 at default.