Computing the 95% confidence interval of a continuous outcome based on STATA's linear regression outputs

[Johannes Mischlinger]

Dear all,

I came across a seemingly simple question that I would like to discuss here.

On the context: The dataset contains antibody concentrations from people with rheumatic disease (exposure variable "rheum_yn"==1) and people without rheumatic disease (exposure variable "rheum_yn"==0). Crude and adjusted linear regressions were performed to predict the mean antibody concentrations in people with rheumatic disease versus people without rheumatic disease. Based on the STATA output, it is easy to calculate the mean antibody concentration following the logic 'outcome = slope*exposure + intercept' (i.e. antibody concentration = -0.153*rheum_yn[0 or 1] + 0.22.) However, what is not intuitive is the question how to compute the 95% confidence interval of the mean antibody concentration.

While for those people without rheumatic disease ("rheum_yn==0") it seems to be the 95% CI of the intercept, I wonder how to compute this for those with rheumatic disease ("rheum_yn==1"). For example to compute the lower 95% CI bound, is it necessary to add together the lower 95% CI bounds of the slope AND the intercept? The same question then applies to the computation of the upper 95% CI bound.

Regards,
Hannes

[Felix Bittmann]

I understand that you want the CI for the both group means.

reg tetanuslog i.rheum_yn
margins rheum_yn