Because of the difficulties involved in actually measuring concentricity error, I'm not sure where your data is coming from. In my experience, in almost every instance I've ever seen where concentricity was specified it was really control of position or runout that was needed. The input values for your formula must be the result of measurement for the median points of diametrically opposed elements of the feature being controlled. In practical terms, this involves using two indicators positioned 180 degrees apart on the controlled feature, and determing the net difference in the readings at different points along the feature, and then identifying the worst case, which is the greatest net difference between any two readings. Thus it appears that perhaps your formula may be as simple as finding the difference between two (perhaps differently signed) values.
If this isn't clear, or if I haven't answered your question, please post back and I'll give it another go.