Hi ROOTers and especially those using RooFit/RooStats,
For a toy model of a real application I was trying to build a composite likelihood using RooFit capabilities. The RooNLLVar class returns a proper Likelihood function though only for one given dataset. If I were to combine likelihoods to form a composite likelihood that can be fed into Minuit, I wonder whether there is anything that I could use.
To illustrate the case:
Suppose you have m-measurements of the same physical phenomenon but
independent experiments. Each experiment can be fitted and returns you a
"best value" for your model and gives an associated likelihood function L_i.
Since all experiments describe the same physics one can combine the results
and get a better constrained fit. In order to do that, one forms the
combined (composite) likelihood
L = Product(L_i)
which depends on the same model parameters as each individual L_i.
I would be very glad if any of you guys could help me figuring out how to do this kind of task. Is there anything implemented in RooFit/RooStats?
Best
Stephan
Received on Thu Sep 23 2010 - 22:43:32 CEST
This archive was generated by hypermail 2.2.0 : Fri Sep 24 2010 - 17:50:01 CEST