MIT’s Richard Lindzen: ‘When data conflicts with models, a small coterie of scientists can be counted upon to modify the data’