How do I sum two or more regression coefficients, including standard errors? I currently do it 'manually', but I suspect that there is a better way. Below is an example of what I would do to find the cumulative coefficient of female, MW, and the interaction. Ideally, a solution would be able to accommodate n>1.
[EDIT:] I figured it out! It's simply:
Code:
#Version 15.1 webuse nhanes2, clear regress bmi female##region mat a1=e(b)' mat b1=e(V) mat c1=(a1,b1) mat list c1 loc beta1=c1[2,1] loc beta2=c1[12,1] loc beta3=c1[4,1] loc err1=(c1[2,3]) loc err2=(c1[12,13]) loc err3=(c1[4,5]) loc cov1=c1[12,3] loc cov2=c1[4,3] loc cov3=c1[12,5] loc cumb=`beta1'+`beta2'+`beta3' loc cumerr=(`err1'+`err2'+`err3'+2*`cov1'+2*`cov2'+2*`cov3')^0.5 loc cumerrl=`cumb'-invnorm(0.975)*`cumerr' loc cumerrh=`cumb'+invnorm(0.975)*`cumerr' dis `beta1' `beta2' `beta3' `err1' `err2' `err3' `cov1' `cov2' `cov3' dis `cumb' `cumerr' dis "Cumulative effect is `cumb' (`cumerrl', `cumerrh')"
Code:
lincom 2.region+1.female+1.female#2.region
