I wrote a model by myself and run it in DYNARE++. Unfortunately, I got negative variance for several variables. They are all close to zero. I guess this happened because of round error in underlying computation. Actually, their variances should be zero. Also, I think I have to do the simulation (dynare_simul) to get the correct variance for variables in my model. Am I correct? please help me!
Yes, variances should either be 0. Due to numerical error, they can be up to plus/minus 10^-8 or so. When simulating series and computing variances, they cannot be negative.