I'm trying to evaluate an ordered model broken through with polr: polr (Rating ~ Currac + Debt + Inflation + GDPpc + GDPgr + Ratio + Levelofdev + Eurozone + Default, method = "broken")
where Rating is an ordered discrete dependent variable and independent variables are a set of economic determinants (for example, inflation).
However, I still get the same error:
Error in optimization (s0, fmin, gmin, method = "BFGS", ...): the initial value in 'vmmin' is not final
I found it in C code (src / main / optim.c), on line 523 in the vmmin procedure. The error is called because the result of the fminfn function is not finite. In this function, fminfn, on line 82, the result is calculated as result = REAL (s) [0] / (OS-> fnscale). It doesn't really matter to me, however.
I realized that this should have something to do with a function evaluating an infinite value at starting values. I was hoping that someone could explain this in more detail or help me overcome this problem.
Excluding / excluding NA values from a dataset does not solve the problem. However, when I remove either inflation or GDPpc from the set of regressors, it suddenly works fine. And when I delete everything else, except for inflation and GDPpc, it also still works.
If you have the same problem or know how to handle this, I would really appreciate your reply.
Thank you very much,
Wouter
source share