[DFTB-Plus-User] Hubbard correction in DFTB

[juhasz.g.aa]

Dear DFTB+ users, 
I would like to ask for advice  DFTB+U method for simulation of transition metal oxides. 
A Hubbard correction for DFT simulation (DFT+U) of transition metals oxides are often used in solid state physics. The U parameter is generally determined from first principle, eg. using the linear-response method of Cococcioni. However I am confused however, how can I do similar corrections in DFTB.

My understanding is that U values for DFTB are assigned in SK files, and unlike in DFT, there are generally non-zero Us assigned for all elements and shells. 
* Should I use new U parameters, if I would like to effect of Hubbard correction (assuming the original SK parameter set was created using DFT without Hubbard correction)?  
* How can I calculate these U parameters? Linear response calculations similar to  Cococcioni.'s method are not available in DFTB+, are they?
* I often see publications where different U values (with same DFT method) tested for reproducing experimental data. Is it a valid approach to similarly test DFTB calculations with various U values keeping other SK parameters the same?

Thank you for your comments.
Best regards,
Gergely Juhasz