After you log-linearize your model and you want to compare the model with data, what is the best approach?

I am asking that because I believe the results are actually deviations from the mean (steady-state).

Can someone please provide more insight into this?

Thanks.