I’m using ROOT version 6.14.04. I am wondering how to plot variables and their correlations through TMVA::DataLoader which used to have the functions TMVA::DataLoader::DrawInputVariable() and TMVA::DataLoader::DrawCorrelationMatrix(), you can check here: http://nbviewer.jupyter.org/github/iml-wg/tmvatutorials/blob/master/TMVAinJupyter_FullTutorial.ipynb if you go to the Visualizing input variables part, but not for this new version.
The functions you refer to are only available to jsmva, the web-interface of TMVA. (Will update this post with a link when I have a good reference for more information on it).
If you want to use a similar functionality in the c++/python version of TMVA you will have to use the TMVAGui. You can start it given your output file from classification/regression like so:
root -l -e 'TMVA::TMVAGui("./path/to/file")'. More information on TMVAGui can be found in the Users’ Guide.
Good to know this info. It looks like many MVAs(RMVA, TMVA, pyMVA) are not unified.