We tune the engine on known surveys.
Then it comes close to them.
That is calibration.Not validation.
The real test comes next.
New questions, out of sample.
Pre-registered, no second chance.
That test has not been run yet.
The engine comes close to the surveys it was tuned on.
That does not prove validity yet.
Fieldwork still has the final word.
Calibrated on known surveys.Not yet validated on new questions.
Space: pause · arrow keys: previous / next scene · click the bar: jump to a scene