The fit of data using a mathematical model is the standard way to know if the model describes data correctly and to obtain parameters that describe the physical processes hidden behind the experimental results. This is usually done by means of a χ2 minimization procedure. Although this procedure is fast and quite reliable for simple models, it has many drawbacks when dealing with complicated problems such as models with many or correlated parameters. We present here a Bayesian method to explore the parameter space guided only by the probability laws underlying the χ2 figure of merit. The presented method does not get stuck in local minima of the χ2 landscape as it usually happens with classical minimization procedures. Moreover correlations between parameters are taken into account in a natural way. Finally, parameters are obtained as probability distribution functions so that all the complexity of the parameter space is shown.