I'm outputting some estimation results to a table for a range of specifications, and producing a long-run estimate of certain variables, which is a function of two previous OLS estimates: say Beta3= Beta2/Beta1. The relevant code below is creating the desired LR coefficients, %coefLR... so far, so good.
Code: Select all
!LRvariables=(!variables/2)-1
for !z=2 to !LRvariables
%coefLR=""
!coefLR1=@round({%pool}.@coefs(2*!z+1)*!decimals)/!decimals
!coefLR2=@round({%pool}.@coefs(2)*!decimals)/!decimals
%coefLR=@str(!coefLR1/!coefLR2)
_result_large(5+!z+!variables+!posi_down,1)=_var(2*!z+1,1)+": LR effect"
_result_large(5+!z+!variables+!posi_down,!position+1)=%coefLR
next
What I'd like to do now is generate standard errors (and as a result, t-statistics) for these new LR coefficients. My econometric theory is a bit rusty but I can't recall any algebraic equation that will get me from the standard errors of the original 2 estimators Beta1 and Beta2, to the standard error of Beta3.
Therefore, I'm looking for a way to directly generate standard errors for the estimations of Beta3. Any ideas?