T-optimality: a stopping rule for a first order algorithm

In the optimal design theory, the T-optimality criterion is useful for the discrimination between two competitive models. This criterion has an interesting statistical interpretation as the power of a test for the fit of a second model when the first one is true. Usually there is not a closed form for the T-optimum design and it must be computed through an iterative procedure. In this short note a stopping rule for a first order algorithm is provided. The proposed stopping rule has the interesting feature that the algorithm will always stop at a design which reaches a minimum fixed efficiency. In other words, the algorithm stops when it reaches a design efficiency (with respect to the unknown T-optimum design) as good as wanted.