first post, hope I’m not breaking any rules
I am approaching bayesian analysis.
My problem is the following:
A model to describe a certain phenomenon was proposed in a paper. The model depends on 4 parameters and these were estimated with a fit to experimental data. I want to update these values.
I have added points to the data base. I could follow the same procedure of the paper (a simple fit), but I think it is more correct to make a Bayesian inference to estimate the most probable values of the parameters.
So I defined a model in pymc that considers the 4 model parameters as Gaussians (using mean and sigma reported in the paper) - my priors - and defined a function that combines these parameters in agreement with the theoretical model.
I added an uncertainty to the experimental points, defining the sigma (e) as HalfCauchy
Then I estimate the value of the parameters
y_pre= pm.Normal(‘y_pre’, mu=u, sigma=e, observed=x_data)
where u is the function that combines the 4 parameters, and x_data the set of experimental points (old and new).
Is it reasonable?
Is it possible to post the full code?