# Estimating a non-linear model

Hello,
I am trying to estimate a non-linear model like the example in the DYNARE user guide. But I am confused on what kind of data to use.
Am I to estimate using the data as they are, for example, GDP in its actual form, because my model is non-linear?
I will be glad if anyone can help clarify this.

What exactly do you mean by this?

@bdombeck
Thank you for replying. This is an example of one of the equations in my model

yF = a k^alpha hF^(1-alpha);
the equation is a non-linear model of output

To estimate, my model using bayesian technique, I am to provide my dataset.
My question is that should my output data (gdp) be in its raw from as gathered from World Bank or I am to do a form of transformation.
I hope this is clearer now. Thanks

Maybe I will try to be more clear about what is unclear from my perspective:

1. You have used the term “actual form” and also “raw from” when referencing the data you are collecting for use in estimation. I do not know what you mean by these two distinct ways of characterizing the data. Perhaps you could be more specific about the type of data you have collected and what the different “forms” you are trying to decide between. At the moment there is no way to know if you mean “level vs growth rate” or “level vs per capita” or “filtered vs unfiltered” or any number of other ways we can think about macro data.

2. Your model is a suggestion for the data generating process. Most DSGE models are inherently nonlinear, but unless there are particular nonlinearities characteristics you are hoping to explore (e.g. the role of risk or stochastic volatility) what you probably want to do is “estimate a first order approximation of the model” which will itself be given as a typical linear state-space model. This is what e.g. Smets and Wouters (2007) do. The phrase “estimate a nonlinear model” can be taken to mean a variety of things, so it would be useful for you to explain your goal in more detail.

@bdombeck