for example, i use a dsge model to describe an economy,and then i modified the monetray policy rules of the model and estimate the model again, how to compare the two model in its describing and explaining the economy, i mean is there a Criteria to judge which model is better or optimal? if i use marginal density to compare, is it siutable?

I think you can use the Geweke’s Modified Harmonic Mean estimator of the marginal likelihood to do comparison. Pls refer to An and Frank (2007) page 146. By Dynare, you can first estimate model with filename.mod, then in matlab load the .mat file filename_results.mat, the code is "load filename_results.mat ", then write “HM=oo.MarginalDensity.ModifiedHarmonicMean”, then you can get the value of Harmonic Mean estimator . About the econnomic sense and how to use it , see An and Frank (2007).