different optimization routines & bad bayesian IRFs plot

[ericlearndynare]

Hi everybody!
i was doing a DSGE model using bayesian technique. when i set in the "estimation()" with "mode_compute = 4", i got note:
??? Error using ==> chol
Matrix must be positive definite.

Error in ==> metropolis_hastings_initialization at 52
d = chol(vv);

Error in ==> random_walk_metropolis_hastings at 58
[ ix2, ilogpo2, ModelName, MhDirectoryName, fblck, fline, npar, nblck, nruns, NewFile, MAX_nruns, d ] = ...

while after using optimization routine "mode_compute = 6", i can get the estimation results, but the Bayesian IFRs are not desirable, most like sharp cliffs or responses in big scale between each plot.

i don't know what to do next. any help is much appreaciated！

eric

[ericlearndynare]

can anyone help me?

[jpfeifer]

If your model and the observation equation is correct, the problem is with finding the mode given your starting values for estimation. This is a general and common problem (lots of posts). One way would be to increase the default parameters of mode_compute=6, see http://www.dynare.org/DynareWiki/MonteCarloOptimization