How can adjust the internal precision of Minuit and the external precision of the types of numbers I use to match these precisions?
I think that the best precision is the double precision. I’d like to be sure that the two precisions match, and I’d like also the best precision, or almost a very good precision.
Minuit uses always double precision and you have also to ensure that the function you are passing to be minimized is evaluated in double precision.
Furheremore in Minuit you can control the tolerance, the point when the minimization will stop. The tolerance is passed as optional parameter to the Migrad command.
The minimization will stop when the estimated distance to the minimum is less than 0.001*tolerance
I hope I have answered your question, otherwise please let me know,