Data simulation

Good Day Everyone,

I am a bit confused on data simulation, I hope someone to clarify please.
I have my DSGE model running, produced IRFS. Now am supposed to look at data simulation. I got the actual data, Normalized it and got the mean, variance, STD and correlation among the variables, i thought this was done. However as i read more on data simulation am supposed to have the actual data and simulated data and compare the two. How do i get this simulated data? And is it about the statistics only? and how best to explain and compare.
Thank you for your help!

1 Like

stoch_simul should provide you with the model moments that you can compare to the data moments. It’s often better to compare theoretical model moments to the data moments than to use simulated data moments (periods>0).

Thank you,
I will do that.

Hello again i have these data moments, the data was log linearised. for household expenditure, investment government expenditure, transfers and lastly its GDP for Normal Data, I wanted to ask is it okey that i have negative mean values?
0.0488 0.1547 0.0813 0.1550 0.1322 std
-0.0901 -0.1885 -0.0937 -0.1813 -0.1377 mean
0.0024 0.0239 0.0066 0.0240 0.0175 variance

I am not sure I can follow. You took the logarithm of the levels. Of course those can be negative.

I had a presentation and that seemed to be a concern but yet its likely to have negative mean and i wondered why. So I tried to take logs first before normalising the data and now i have positive mean.
Thank you
Now I have another question: When comparing real and simulated data, is it necessary for them to match perfectly? Specifically, should the moments, such as correlation coefficients, be identical in both datasets

  1. I still don’t understand. You cannot normalize after taking logs.
  2. No, they will never be identical, given that a model is a stylized depiction of reality. But they should not differ too much. Otherwise, the model is too misspecified.