I have a question related to obtaining the variance matrix after a fit. The variance matrix can be obtained using TH1::Fit, and the fit option “S”. The terms corresponding to fixed parameters will be zero.
In my case, most of the parameters are kept fixed during the fit.
For the error propagation, I was using a sampling MonteCarlo method (consisting in sampling the fixed parameters according to their uncertainties, and accumulating the fitted parameters, so the std dev will give the error).
I was assuming the fixed parameters are not correlated, which may not be true.
Question: given a TF1, and a set of parameters (free & fixed), is there a way of computing the correlation matrix for the fixed parameters? ( I don’t know if converting the fixed parameters into free parameters tightly bounded would produce that result )
( If the question is too basic, a bibliographic source would be also helpful. )
Thank you for your time.
ROOT Version: ROOT 6.10/08