Could you explain to me how using the PiG model you calculate joint centres at the hip and the knee. I am assuming that at the hip you use a known regression equation (Davis, Bell, Harrington). At the knee you are required to normally use the knee width which is measured clinically in order to define the medial-lateral position of the joint centre. I do not think Anybody does this as I cant remember inputting a knee width value, so I would like to know how you calculate the joint centres?

I think you have the wrong conception of what AnyBody does when scaling the models. In fact, the model is only scaled linearly and some marker positions are optimized. The concept is different from what you see in traditional gait analysis where the model is completely defined by the markers. It is conceptually different. A good explanation of what AnyBody does is given by Andersen: