[michaelt]

  [2] is similarly vague, [...] isn't "contact
  with the instructor" significant?

Well, they did compare online courses.

The article doesn't detail how their online courses are structured, but when I've done free udacity/coursera/edx courses, contact with the instructor has been nonexistent.

Seems to me it'd be very difficult to design a test to investigate this that wouldn't have some valid methodological criticisms. Even if you had lecturers, delivering online lectures with fake names, using voice-changing software and online-only office hours, you could still criticise that as being unrepresentative of real college lecturing, and having confounding factors if the courses ran at different schools, years, or times.