Page 1 of 1

Problems with Markov Switching Estimation

Posted: Sun May 28, 2017 10:20 am
by loxo
Hello everyone, I've been trying to estimate a two state markov switching model where only the mean changes, but when I try to estimate my equation I get this error (I'm using Eviews 9):


Dependent Variable: LOGOUTPUTT
Method: Markov Switching Regression (OPG - BHHH / Marquardt steps)
Date: 05/28/17 Time: 14:16
Sample (adjusted): 1980Q2 2010Q4
Included observations: 123 after adjustments
Number of states: 2
Initial probabilities obtained from ergodic solution
Standard errors & covariance computed using observed Hessian
Random search: 25 starting values with 10 iterations using 1 standard
deviation (rng=kn, seed=798239572)
Optimization failed: unable to evaluate objective at optimization starting
values
WARNING: Singular covariance - coefficients are not unique

Variable Coefficient Std. Error z-Statistic Prob.

Regime 1

C 0.008562 NA NA NA

Regime 2

C 0.008562 NA NA NA

Common

AR(1) 0.000000 NA NA NA
AR(2) 0.000000 NA NA NA
AR(5) 0.000000 NA NA NA
AR(6) 0.000000 NA NA NA
AR(7) 0.000000 NA NA NA
AR(8) 0.000000 NA NA NA
LOG(SIGMA) -1.839441 NA NA NA

Transition Matrix Parameters

P11-C 0.000000 NA NA NA
P21-C 0.000000 NA NA NA

Mean dependent var 0.006285 S.D. dependent var 0.025251

Does anyone know how to solve this problem?
Thank you very much for your help

Re: Problems with Markov Switching Estimation

Posted: Wed Jun 21, 2017 10:55 am
by EViews Glenn
You need to try more starting values.