Dear Stata-list visitors!
I have a question about how to turn to the coefficient matrix e(b). I use a max-likelihood model that uses explanatory variables b1, b2 and b3 not only in the command ml model, but also inside the model, in intermediate calculations. The model has a variable scale (among others), where scale = b1+b2+b3. So, these three coefficients are the same coefficients, the model is made to estimate for. And I need them (their initial values) already in the start of the first iteration.

I tested two variants to enter the values of explanatory variables, where b1 = `b '[1.1] and b1 = _b [b1] as an example. For second version I used another similar (but simpler) model, which gives me a matrix e(b) for initial values. But I was told that the second version is incorrect, because in the second case these coefficients stay constant, until to the last iteration. These variants produce very different results, with the result of the second variant being more realistic.

The variables b1, b2 and b3 should be coded so that they are the same variables as in the coefficient matrix e(b) and in the command ml model and ml init. And these variables need initial values to make the model work at all. Could anyone recommend the appropriate code?