Problem of "initial values not feasible" in a Markov Switching Model

[Mateus Muller]

Hi, i'm new to Stata and new to MS models.

For my master's dissertation i'm studying alternative taylor's rule models for central bank's reaction function.

I'm trying to run "mswitch dr it L1.it, switch(d.Djt L1.hiato L1.cambioN) states(2) varswitch",
where it=interest rate, Djt=inflation gap, hiato=GDP gap, cambioN=nominal exchange rate change, all monthly data and stationary (Djt stationary at first difference), from 2003m1 to 2017m12.

And i get this error:
Performing EM optimization:

initial values not feasible
r(1400);

I tried to adapt the advice of https://www.stata.com/statalist/arch.../msg00650.html to a mswitch, but it didn't solved my problem (or i did it wrong).

How can i check/estimate the initial values i must use or how do i correct the problem?

Any help will be appreciated.

Thanks,
Mateus

[Phil Bromiley]

You didn't get a quick answer. You'll increase your changes of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex.

It is far more likely that someone can solve a data problem if you provide the data and code than just from your code. I'd start by replicating the examples in the documentation and then try moving a step at a time to your data and model. If you really want to, there are a variety of options listed under maximize in the pdf documentation that you can look into. You could also try giving it initial values.