(How) can we produce a test that shows that the program as a whole works? One thing is that we could show that we get the "correct" (whatsoever this means) max. likelihood point estimate. But can we also show that the likelihoods mapped out by the code are reasonable? Potentially on a simplified result?
I'm not sure whether this is too ambitious, but it'd be great if we could show that the code does not only produce numbers, but that they are meaningful (even if potentially only for simplified results).
(How) can we produce a test that shows that the program as a whole works? One thing is that we could show that we get the "correct" (whatsoever this means) max. likelihood point estimate. But can we also show that the likelihoods mapped out by the code are reasonable? Potentially on a simplified result?
I'm not sure whether this is too ambitious, but it'd be great if we could show that the code does not only produce numbers, but that they are meaningful (even if potentially only for simplified results).