Sorry for the delay. I thought that we were finished!

My previous post shows the formula with a common **B** and **beta** and conditional parameters on **Category** levels. What else do you need? The script in my first reply automates building such a formula. I can't understand what else you might require from a model stand-point.

Are you asking about how the **Nonlinear** platform solves such a problem (estimate the parameters)? It uses a *numerical optimization procedure* and a *loss function* to monitor. It uses the given *starting values* and then varies them to improve the result of the loss function (minimization). The default loss function is *sum of the squared errors*, or *least squares*. It monitors several measures of change to determine when *convergence* is achieved and it stops.

Learn it once, use it forever!