As you have stated the problem, you should rewrite the model so that
it specified in terms of the parameters: beta, gamma, eta_j and
lambda where lambda is the constant ratio of delta_j/eta_j and then
use -nl- for non-linear least squares. It is always best to
substitute out constraints if you can do that.
However, I suspect that you might have something else in mind. Do
you really want to test whether the ratios delta_j/eta_j are constant
over j? In that case, your question is not really about constraints
but about testing a set of non-linear hypotheses and you need to
consider -testnl- or other test procedures.
