Catherine Goehring, Arnaud Perrier and Alfredo Morabia published a paper in

Statist. Med. 2004; 23:125–135 and provided a useful way of computing variance of spectrum bias for raio of {(LR+ subgroup)/ (LR+ all)} [see screenshots]. It works!

However, the formula for computing variance for spectrum bias using {(LR- subgroup)/ (LR- all)} is not provided. I have emailed Dr Goehring and Professor Morabia about this. Meanwhile, can someone help please?

Thanks.