I am a beginner in using dynare and although I started to read a lot about this subject, I encountered some misunderstandings:

  1. First of all, supposing I want to estimate a model through bayesian techniques (using the metropolis-hastings algorithm), with observable variables like GDP, consumption, investment, interest rate, inflation, hours worked, wages. If I introduce the equation in Dynare for consumption for example:
    C = h/(1+h)*C(-1) + 1/(1+h)C(+1)+(sigma_c - 1)/((1+lambda_w)(1+h)*sigma_c) * (L - L(+1)) - (1-h)/((1+h)*sigma_c) * Rauxil(-1) + (1-h)/((1+h)*sigma_c) *eps_B;

how should my actual data look like? I am confused as the data which I have extracted (quarterly data, seasonally adjusted) is not stationary, presenting an increasing trend and I don’t know if I should stationarize it and afterwards modify also the code in Dynare somehow. And if I should stationarize it, detrending it, how should I do that? And also, what about the interest rate and inflation which are presented as %? Shouldn’t I express all the variables in the model in the same way?

  1. About the calibrated parameters: in general, in many articles that I have read, these appear to be standard values. For example, for beta (the discount factor) almost in every cases the calibrated value is 0.99 or for capital depreciation rate is 0.025. It is ok to use the same calibration for different type of economies? And if you state other values, based on what should you do that?

  2. I understood that the initval for the observable values should be as close to the ones at the steady state. How can I better aproximate that? Which way do I know that my values are correctly posted in the initval subsection?

Thank you in advance! I would really appreciate your help!

1.) Please see Pfeifer(2013): “A Guide to Specifying Observation Equations for the Estimation of DSGE Models” sites.google.com/site/pfeiferecon/Pfeifer_2013_Observation_Equations.pdf. Short answer is: use first differences for non-stationary variables. For interest rates, there are several ways detailed in the document. The important thing is that the data matches the model variables.
2.) No, calibration must be tailored to the respective economies. Usually, calibration means you pick some parameters to match observed ratios in the data. See for example sites.google.com/site/pfeiferecon/RBC_news_shock_model.mod for a mod-file where delta and beta are set to match the specific investment-output and capital-output ratio of the studies economy.
3.) Put the resid(1); command before steady to see how good your starting values are. All residuals should be as close to 0 as possible. Even better for estimation: use a steady state file. See the respective remark in the linked document.