Hello. Is there a way to fix residue unexplained variability (sigma) to zero in FOCE?
It’s a parameter like any other so you can restrict it with the
constantcoef keyword argument. However, restricting it to zero doesn’t really make sense for a statistical model and it would make the almost certainly make the optimization fail since the resulting likelihood would be infinite for almost all parameter values.
Instead of fixing your sigma to 0, you can try with a lower value. you can fix the value using the code below.
fit(Model, data, Param, Pumas.FOCE(), constantcoef=(σ_add =0.0001,))