New attempt with Optim.jl v2.2

[juliohm]

cc: @pkofod

[juliohm]

I've added a small delta to the lower bound of the box constraints to avoid the corner case you mentioned (all proportions equal to zero). Unfortunately, it produces a new error with LBFGSB.

[pkofod]

Oh, I had not seen this when I did #86

[pkofod]

There is nothing special about the box constraints in #85, yet we still get test failures with Optim.jl v2.x, which seems to indicate an issue on the optimization side. Can we try to debug it further over there?

I don't think I agree. From my own debugging the problems I saw and described were not related to issues with LBFGSB, it was related to the facts I described for example that proportions could not be evaluated at a zero vector. I see you added a small lower bound here, and now something else fails, but I don't know what those tests signify. I think you'd need to look at the fitted proportions to see why those tests now fail.