CVM for dataall.txt, data source: http://www.nature.com/nature/journal/v430/n6995/extref/nature02478-s1.htm




Calibration R=0.23, RE=-0.55, 1-sigma=0.22. Not good. But if we sort proxies according to correlation with instrumental, and take 18 best, we'll do better:


Calibration R=0.61, RE=0.22, 1-sigma=0.16.
I would use Classical Calibration Estimator (CCE, A2 before scaling step):


This is optimal in many ways, but R=0.23, RE=-18 and 1-sigma 0.78